{"v":1,"built":"2026-09-22T02:04:59.033Z","speed":"221 of 225 models measured","providers":[{"name":"Cerebras","slug":"cerebras","hq":"US","privacy":"https://www.cerebras.ai/privacy-policy","terms":"https://www.cerebras.ai/terms-of-service","status":"https://status.cerebras.ai/"},{"name":"Morph","slug":"morph","hq":"US","dc":["US"],"privacy":"https://morphllm.com/privacy","terms":"https://www.morphllm.com/privacy/tos"},{"name":"StepFun","slug":"stepfun","terms":"https://platform.stepfun.ai/docs/en/agreement/userservice"},{"name":"Deepgram","slug":"deepgram","privacy":"https://deepgram.com/privacy","terms":"https://deepgram.com/terms"},{"name":"VoyageAI by MongoDB","slug":"voyageai","privacy":"https://www.voyageai.com/privacy","terms":"https://www.voyageai.com/tos"},{"name":"Black Forest Labs","slug":"black-forest-labs","privacy":"https://bfl.ai/legal/privacy-policy","terms":"https://bfl.ai/legal/terms-of-service"},{"name":"Moonshot AI","slug":"moonshotai","hq":"SG","dc":["SG"],"privacy":"https://platform.moonshot.ai/docs/agreement/userprivacy","terms":"https://platform.moonshot.ai/docs/agreement/modeluse"},{"name":"Stealth","slug":"stealth","terms":"https://openrouter.ai/terms/stealth"},{"name":"Reka","slug":"reka","privacy":"https://reka.ai/legal/privacy-policy","terms":"https://reka.ai/legal/terms-of-use"},{"name":"Arcee AI","slug":"arcee-ai","hq":"US","dc":["US"]},{"name":"SambaNova","slug":"sambanova","hq":"US","privacy":"https://sambanova.ai/privacy-policy","terms":"https://sambanova.ai/terms-and-conditions","status":"https://status.sambanova.ai/"},{"name":"AI21","slug":"ai21","hq":"IL","privacy":"https://www.ai21.com/privacy-policy/","terms":"https://www.ai21.com/terms-of-service/","status":"https://status.ai21.com/"},{"name":"Minimax","slug":"minimax","hq":"SG","dc":["US"],"privacy":"https://www.minimax.io/platform/protocol/privacy-policy","terms":"https://www.minimax.io/platform/protocol/terms-of-service"},{"name":"Io Net","slug":"io-net","hq":"US","privacy":"https://io.net/privacy","terms":"https://io.net/terms","status":"https://status.io.net/"},{"name":"Claude Platform on AWS","slug":"claude-on-aws","hq":"US","privacy":"https://www.anthropic.com/legal/privacy","terms":"https://www.anthropic.com/legal/commercial-terms","status":"https://status.anthropic.com/"},{"name":"Ionstream","slug":"ionstream","hq":"US","dc":["US"],"privacy":"https://ionstream.ai/privacy-policy/","terms":"https://ionstream.ai/terms-and-conditions/"},{"name":"Cohere","slug":"cohere","hq":"US","privacy":"https://cohere.com/privacy","terms":"https://cohere.com/terms-of-use","status":"https://status.cohere.ai/"},{"name":"Z.AI","slug":"z-ai","hq":"SG","dc":["SG"],"privacy":"https://chat.z.ai/legal-agreement/privacy-policy","terms":"https://chat.z.ai/legal-agreement/terms-of-service"},{"name":"BaseTen","slug":"baseten","hq":"US","privacy":"https://www.baseten.co/privacy-policy","terms":"https://www.baseten.co/terms-and-conditions","status":"https://status.baseten.co/"},{"name":"Groq","slug":"groq","hq":"US","privacy":"https://groq.com/privacy-policy/","terms":"https://groq.com/terms-of-use/","status":"https://status.groq.com/"},{"name":"Modal","slug":"modal","hq":"US","privacy":"https://modal.com/legal/privacy-policy","terms":"https://modal.com/legal/terms","status":"https://status.modal.com/"},{"name":"DeepSeek","slug":"deepseek","hq":"CN","privacy":"https://chat.deepseek.com/downloads/DeepSeek%20Privacy%20Policy.html","terms":"https://chat.deepseek.com/downloads/DeepSeek%20Terms%20of%20Use.html","status":"https://status.deepseek.com/"},{"name":"Crusoe","slug":"crusoe","hq":"US","privacy":"https://legal.crusoe.ai/open-router#cloud-privacy-policy","terms":"https://legal.crusoe.ai/open-router#managed-inference-tos-open-router","status":"https://status.crusoecloud.com/"},{"name":"Darkbloom","slug":"darkbloom","privacy":"https://www.darkbloom.dev/privacy.html","terms":"https://www.darkbloom.dev/terms.html"},{"name":"Mancer 2","slug":"mancer","privacy":"https://mancer.tech/privacy","terms":"https://mancer.tech/terms"},{"name":"NextBit","slug":"nextbit","hq":"ES","dc":["ES"],"privacy":"https://www.nextbit256.com/docs/privacy-policy","terms":"https://www.nextbit256.com/docs/terms-of-service"},{"name":"Infermatic","slug":"infermatic","privacy":"https://infermatic.ai/privacy-policy/","terms":"https://infermatic.ai/terms-and-conditions/"},{"name":"Avian","slug":"avian","hq":"US","privacy":"https://avian.io/privacy","terms":"https://avian.io/terms"},{"name":"Inflection","slug":"inflection","privacy":"https://inflection.ai/privacy-policy","terms":"https://developers.inflection.ai/tos"},{"name":"xAI","slug":"xai","hq":"US","privacy":"https://x.ai/legal/privacy-policy","terms":"https://x.ai/legal/terms-of-service-enterprise","status":"https://status.x.ai/"},{"name":"Switchpoint","slug":"switchpoint","hq":"US","dc":["US"]},{"name":"Sail Research","slug":"sail-research","hq":"US","privacy":"https://www.sailresearch.com/privacy","terms":"https://www.sailresearch.com/terms","status":"https://status.sailresearch.com/"},{"name":"Ollama","slug":"ollama"},{"name":"Tencent","slug":"tencent","hq":"CN","dc":["SG"],"privacy":"https://www.tencentcloud.com/en/document/product/1300/78952","terms":"https://www.tencentcloud.com/en/document/product/301/78869","status":"https://status.tencentcloud.com"},{"name":"DeepInfra","slug":"deepinfra","hq":"US","privacy":"https://deepinfra.com/privacy","terms":"https://deepinfra.com/terms","status":"https://status.deepinfra.com/"},{"name":"Amazon Nova","slug":"amazon-nova","hq":"US","dc":["US"]},{"name":"Cosine","slug":"cosine"},{"name":"Mara","slug":"mara","hq":"US","dc":["US"],"privacy":"https://www.mara.com/mara-legals/ai-policies#cloud-pp","terms":"https://www.mara.com/mara-legals/ai-policies#cloud-toc"},{"name":"Perceptron","slug":"perceptron","hq":"US","privacy":"https://www.perceptron.inc/privacy-policy","terms":"https://www.perceptron.inc/terms-of-use"},{"name":"Sourceful","slug":"sourceful","privacy":"https://www.sourceful.com/legal/privacy-policy","terms":"https://www.sourceful.com/legal/spring-terms-of-use"},{"name":"Fish Audio","slug":"fish-audio","privacy":"https://fish.audio/privacy","terms":"https://fish.audio/terms"},{"name":"Azure","slug":"azure","hq":"US","privacy":"https://www.microsoft.com/en-us/privacy/privacystatement","terms":"https://www.microsoft.com/en-us/legal/terms-of-use?oneroute=true","status":"https://status.azure.com/"},{"name":"Krea","slug":"krea","hq":"US","privacy":"https://www.krea.ai/privacy-policy","terms":"https://www.krea.ai/terms"},{"name":"Cloudflare","slug":"cloudflare","hq":"US","privacy":"https://developers.cloudflare.com/workers-ai/privacy","terms":"https://www.cloudflare.com/service-specific-terms-developer-platform/#developer-platform-terms","status":"https://www.cloudflarestatus.com/"},{"name":"Crucible","slug":"crucible"},{"name":"Nebius","slug":"nebius","hq":"NL","privacy":"https://docs.nebius.com/legal/studio/privacy/","terms":"https://docs.nebius.com/legal/studio/terms-of-use/"},{"name":"AionLabs","slug":"aion-labs","hq":"IL","privacy":"https://www.aionlabs.ai/privacy-policy/","terms":"https://www.aionlabs.ai/terms/"},{"name":"Liquid","slug":"liquid","privacy":"https://www.liquid.ai/privacy-policy","terms":"https://www.liquid.ai/terms-conditions"},{"name":"Decart","slug":"decart","hq":"US","dc":["US"],"privacy":"https://cogito.decart.ai/legal/privacy","terms":"https://cogito.decart.ai/legal/terms"},{"name":"Recraft","slug":"recraft"},{"name":"Meta","slug":"meta","hq":"US","privacy":"https://www.facebook.com/privacy/policy/","terms":"https://ai.developer.meta.com/legal/terms-of-service"},{"name":"Phala","slug":"phala","hq":"US","privacy":"https://redpill.ai/privacy","terms":"https://redpill.ai/terms","status":"https://status.phala.network/"},{"name":"Seed","slug":"seed","hq":"SG","privacy":"https://docs.byteplus.com/en/docs/legal/docs-privacy-policy","terms":"https://docs.byteplus.com/en/docs/legal/docs-terms-of-service"},{"name":"CoreWeave","slug":"coreweave","hq":"US","dc":["US"],"privacy":"https://docs.coreweave.com/policies/terms-of-service/privacy-policy","terms":"https://docs.coreweave.com/policies/terms-of-service"},{"name":"SiliconFlow","slug":"siliconflow","hq":"SG","dc":["US"],"privacy":"https://docs.siliconflow.com/en/legals/privacy-policy","terms":"https://docs.siliconflow.com/en/legals/terms-of-service"},{"name":"DekaLLM","slug":"dekallm","hq":"ID","dc":["ID"],"privacy":"https://www.cloudeka.id/id/privacy-policy/","terms":"https://docs.cloudeka.ai/service-portal-ai/end-user-license-agreement"},{"name":"Poolside","slug":"poolside","privacy":"https://poolside.ai/legal/privacy","terms":"http://poolside.ai/legal"},{"name":"ModelRun","slug":"modelrun","hq":"US","dc":["US"],"privacy":"https://www.modular.com/legal/privacy","terms":"https://www.modular.com/legal/terms"},{"name":"FakeProvider","slug":"fake-provider","hq":"US"},{"name":"Modular","slug":"modular","hq":"US","dc":["US"],"privacy":"https://www.runmodelrun.com/privacy-policy.html","terms":"https://www.runmodelrun.com/TOS.html"},{"name":"Cirrascale","slug":"cirrascale","privacy":"https://www.cirrascale.com/privacy-policy","terms":"https://www.cirrascale.com/terms-of-service"},{"name":"Ambient","slug":"ambient"},{"name":"Anthropic","slug":"anthropic","hq":"US","privacy":"https://www.anthropic.com/legal/privacy","terms":"https://www.anthropic.com/legal/commercial-terms","status":"https://status.anthropic.com/"},{"name":"Baidu","slug":"baidu","hq":"CN","privacy":"https://intl.cloud.baidu.com/en/doc/Agreements/s/Plr0fi68q-intl-en","terms":"https://intl.cloud.baidu.com/en/doc/Agreements/s/bmesahnjh-intl-en"},{"name":"AtlasCloud","slug":"atlas-cloud","hq":"US","privacy":"https://www.atlascloud.ai/privacy"},{"name":"Together","slug":"together","hq":"US","privacy":"https://www.together.ai/privacy","terms":"https://www.together.ai/terms-of-service","status":"https://status.together.ai/"},{"name":"Runway","slug":"runway","hq":"US","privacy":"https://runwayml.com/privacy-policy","terms":"https://runwayml.com/terms-of-use","status":"https://status.runwayml.com"},{"name":"Nex AGI","slug":"nex-agi","hq":"CN","dc":["SG"],"privacy":"https://api.nex-agi.cn/privacy-policy"},{"name":"DigitalOcean","slug":"digitalocean","privacy":"https://www.digitalocean.com/legal/privacy-policy","terms":"https://www.digitalocean.com/legal/terms-of-service-agreement"},{"name":"Thinking Machines","slug":"thinkingmachines","privacy":"https://thinkingmachines.ai/legal/tml-free-research-api-tier-privacy-notice.pdf","terms":"https://thinkingmachines.ai/legal/tml-free-research-api-tier-terms-of-service.pdf"},{"name":"Chutes","slug":"chutes","hq":"US","privacy":"https://chutes.ai/privacy","terms":"https://chutes.ai/tos"},{"name":"Inceptron","slug":"inceptron","hq":"SE","dc":["FI"],"privacy":"https://www.inceptron.io/privacy","terms":"https://www.inceptron.io/termsofservice"},{"name":"Upstage","slug":"upstage","privacy":"https://www.upstage.ai/privacy-policy","terms":"https://www.upstage.ai/terms-of-service"},{"name":"Google","slug":"google-vertex","hq":"US","privacy":"https://cloud.google.com/terms/cloud-privacy-notice","terms":"https://cloud.google.com/terms/","status":"https://status.cloud.google.com/products/sdXM79fz1FS6ekNpu37K/history"},{"name":"Xiaomi","slug":"xiaomi","hq":"CN","dc":["SG","NL"],"privacy":"https://platform.xiaomimimo.com/#/docs/terms/privacy-policy","terms":"https://platform.xiaomimimo.com/#/docs/terms/user-agreement"},{"name":"Google AI Studio","slug":"google-ai-studio","hq":"US","privacy":"https://cloud.google.com/terms/cloud-privacy-notice","terms":"https://cloud.google.com/terms/"},{"name":"Parasail","slug":"parasail","hq":"US","privacy":"https://www.parasail.io/legal/privacy-policy","terms":"https://parasail.io/legal/terms-of-service"},{"name":"Databricks","slug":"databricks","hq":"US","privacy":"https://www.databricks.com/legal/privacynotice","terms":"https://www.databricks.com/terms-of-use","status":"https://status.databricks.com"},{"name":"Nvidia","slug":"nvidia","hq":"US","dc":["US"],"privacy":"https://www.nvidia.com/en-us/about-nvidia/privacy-policy/","terms":"https://assets.ngc.nvidia.com/products/api-catalog/legal/NVIDIA%20API%20Trial%20Terms%20of%20Service.pdf"},{"name":"Featherless","slug":"featherless","hq":"US","privacy":"https://featherless.ai/privacy","terms":"https://featherless.ai/terms","status":"https://featherless.ai/status"},{"name":"Makora","slug":"makora","status":"https://makora.statuspage.io"},{"name":"Relace","slug":"relace","privacy":"https://www.relace.ai/privacy-policy","terms":"https://www.relace.ai/terms-of-use","status":"https://status.relace.ai"},{"name":"OpenAI","slug":"openai","hq":"US","privacy":"https://openai.com/policies/privacy-policy/","terms":"https://openai.com/policies/row-terms-of-use/","status":"https://status.openai.com/"},{"name":"InferenceNet","slug":"inference-net","hq":"US","privacy":"https://inference.net/privacy-policy","terms":"https://inference.net/terms-of-service","status":"https://status.inference.net"},{"name":"Alibaba","slug":"alibaba","hq":"SG","dc":["SG","CN"],"privacy":"https://www.alibabacloud.com/help/en/legal/latest/alibaba-cloud-international-website-privacy-policy","terms":"https://www.alibabacloud.com/help/en/legal/latest/alibaba-cloud-international-website-product-terms-of-service-v-3-8-0","status":"https://status.alibabacloud.com/"},{"name":"AssemblyAI","slug":"assemblyai","dc":["US"],"privacy":"https://www.assemblyai.com/legal/privacy-policy","terms":"https://www.assemblyai.com/legal/terms-of-service","status":"https://status.assemblyai.com"},{"name":"Amazon Bedrock","slug":"amazon-bedrock","hq":"US","privacy":"https://aws.amazon.com/privacy","terms":"https://aws.amazon.com/service-terms/","status":"https://health.aws.amazon.com/health/status"},{"name":"Clarifai","slug":"clarifai","hq":"US","dc":["US"],"privacy":"https://www.clarifai.com/company/privacy-policy","terms":"https://www.clarifai.com/company/terms"},{"name":"Quiver","slug":"quiver"},{"name":"TypeSafe","slug":"typesafe","privacy":"https://typesafe.ai/legal/privacy-policy","terms":"https://typesafe.ai/legal/terms"},{"name":"Near AI","slug":"near-ai","privacy":"https://near.ai/privacy-policy","terms":"https://near.ai/terms-of-service"},{"name":"StreamLake","slug":"streamlake","hq":"CN","privacy":"https://www.streamlake.ai/document/DOC/mgkci47q13qr66h9i54","terms":"https://www.streamlake.ai/document/DOC/mgkchnd89grpt1961fw","status":"https://status.streamlake.ai/"},{"name":"Venice","slug":"venice","hq":"US","dc":["US"],"privacy":"https://venice.ai/legal/privacy-policy","terms":"https://venice.ai/legal/tos"},{"name":"Sakana AI","slug":"sakana","privacy":"https://console.sakana.ai/privacy-policy","terms":"https://console.sakana.ai/terms-of-service"},{"name":"Inception","slug":"inception","privacy":"https://www.inceptionlabs.ai/terms#privacy-policy","terms":"https://www.inceptionlabs.ai/terms"},{"name":"Inferact vLLM","slug":"inferact-vllm"},{"name":"Friendli","slug":"friendli","hq":"US","privacy":"https://friendli.ai/privacypolicy","terms":"https://friendli.ai/terms-of-service","status":"https://status.friendli.ai/"},{"name":"Unbiased","slug":"unbiased","privacy":"https://unbiased.ai/terms/#privacy-policy","terms":"https://unbiased.ai/terms/"},{"name":"AkashML","slug":"akashml","privacy":"https://akashml.com/privacy","terms":"https://akashml.com/terms"},{"name":"HeyGen","slug":"heygen","hq":"US","privacy":"https://www.heygen.com/privacy","terms":"https://www.heygen.com/terms"},{"name":"Perplexity","slug":"perplexity","hq":"US","privacy":"https://www.perplexity.ai/hub/legal/privacy-policy","terms":"https://www.perplexity.ai/hub/legal/perplexity-api-terms-of-service","status":"https://status.perplexity.ai/"},{"name":"Tenstorrent","slug":"tenstorrent","hq":"US","dc":["US"],"privacy":"https://tenstorrent.com/privacy","terms":"https://tenstorrent.com/terms"},{"name":"Wafer","slug":"wafer","hq":"US","privacy":"https://www.wafer.ai/privacy-policy","terms":"https://www.wafer.ai/terms"},{"name":"OpenInference","slug":"open-inference","hq":"US","privacy":"https://www.openinference.ai/privacy","terms":"https://www.openinference.ai/terms"},{"name":"Fireworks","slug":"fireworks","hq":"US","privacy":"https://fireworks.ai/privacy-policy","terms":"https://fireworks.ai/terms-of-service","status":"https://status.fireworks.ai/"},{"name":"GMICloud","slug":"gmicloud","hq":"US","dc":["US"],"privacy":"https://docs.gmicloud.ai/privacy","terms":"https://www.gmicloud.ai/terms-and-conditions"},{"name":"Mistral","slug":"mistral","hq":"FR","privacy":"https://mistral.ai/terms/#privacy-policy","terms":"https://mistral.ai/terms/#terms-of-use","status":"https://status.mistral.ai/"},{"name":"PrimeIntellect","slug":"primeintellect"},{"name":"Novita","slug":"novita","hq":"US","privacy":"https://novita.ai/legal/privacy-policy","terms":"https://novita.ai/legal/terms-of-service","status":"https://status.novita.ai/"}],"prose":{"baseline":{"slop":6.898569416278464,"trigram":0.09394566577031677,"notXButY":0.044360602915693347,"repetition":3.934313746974778,"mattr":0.5065404910875029,"grade":7.800271730184228,"sentence":14.590292305292788,"words":17244010},"asOf":"2025-11-07"},"sources":[{"id":"bfcl","name":"Berkeley Function Calling Leaderboard V4","what":"Tool use, multi-turn interaction, search and memory","url":"https://gorilla.cs.berkeley.edu/leaderboard.html","data":"https://gorilla.cs.berkeley.edu/data_overall.csv","licence":"Public leaderboard","ok":true,"ms":304,"asOf":"2026-04-12","n":109,"matched":38},{"id":"eqbench","name":"EQ-Bench","what":"Emotional intelligence, creative writing and longform writing","url":"https://eqbench.com/index.html","data":"https://eqbench.com/creative_writing.js","licence":"Public leaderboard","ok":true,"ms":614,"asOf":null,"n":295,"matched":116},{"id":"openrouter","name":"OpenRouter","what":"Catalog, pricing, context, capabilities, providers","url":"https://openrouter.ai/models","licence":"Public API","ok":true,"ms":4967,"asOf":null,"n":513,"matched":432},{"id":"providers","name":"OpenRouter providers","what":"Who serves each model: headquarters, datacentres, policy links","url":"https://openrouter.ai/api/v1/providers","licence":"Public API","ok":true,"ms":137,"asOf":null,"n":109,"matched":432},{"id":"arena","name":"LM Arena","what":"Human-preference Elo: chat, web dev, vision, documents, search","url":"https://arena.ai/leaderboard","licence":"Public leaderboard","ok":true,"ms":1228,"asOf":null,"n":838,"matched":165},{"id":"epoch","name":"Epoch AI","what":"GPQA Diamond, FrontierMath, SimpleQA Verified, SWE-Bench Verified, AIME","url":"https://epoch.ai/benchmarks","licence":"CC BY","data":"https://epoch.ai/data/benchmarks.csv","ok":true,"ms":588,"asOf":"2026-09-18","n":576,"matched":88},{"id":"terminalbench","name":"Terminal-Bench","what":"Agentic terminal tasks, scored by test suites","url":"https://www.tbench.ai/","licence":"Public leaderboard","ok":true,"ms":247,"asOf":"2026-09-21","n":27,"matched":15},{"id":"swebench","name":"SWE-bench Verified","what":"Real GitHub issues, agent and model together","url":"https://www.swebench.com/","licence":"Public leaderboard","ok":true,"ms":606,"asOf":"2026-02-26","n":180,"matched":38},{"id":"slop","name":"Slop-Score","what":"Stock phrases, repetition and vocabulary spread in generated prose, against a human baseline","url":"https://github.com/sam-paech/slop-score","licence":"MIT, published results","ok":true,"ms":891,"asOf":"2025-11-07","n":22,"matched":15}],"models":[{"id":"anthropic/claude-opus-4.6","name":"Anthropic: Claude Opus 4.6","org":"Anthropic","slug":"claude-opus-4.6","matchKey":"claudeopus46","blurb":"Opus 4.6 is Anthropic’s strongest model for coding and long-running professional tasks.","created":1770219050,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":25,"webSearch":0.01,"alt":{"batch":{"in":2.5,"out":12.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","high","medium","low"]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"tps":35,"ttft":2022,"reqs":5670},{"name":"Azure","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.2,"noSchema":1,"tps":32,"ttft":2193,"reqs":59},{"name":"Google","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"noSchema":1},{"name":"Amazon Bedrock","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.5,"tps":33,"ttft":2065,"reqs":475},{"name":"Anthropic","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.3,"tps":29,"ttft":2345,"reqs":630},{"name":"Google","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"noSchema":1,"tps":34,"ttft":2242,"reqs":342}],"scores":{"arenaChat":{"v":1504.5596374157553,"ci":3.5,"rank":2,"votes":71993,"as":"claude-opus-4-6-high","effort":"high"},"arenaWebdev":{"v":1546.618584729505,"ci":5.6,"rank":28,"votes":18332,"as":"claude-opus-4-6-high","effort":"high"},"arenaVision":{"v":1298.859996739682,"ci":6.7,"rank":5,"votes":20835,"as":"claude-opus-4-6-high","effort":"high"},"arenaDocument":{"v":1507.3187845894272,"ci":6.4,"rank":3,"votes":27929,"as":"claude-opus-4-6-high","effort":"high"},"arenaImageCode":{"v":1540.657027850234,"ci":9.2,"rank":18,"votes":7483,"as":"claude-opus-4-6","effort":""},"arenaSearch":{"v":1253.415899373882,"ci":5,"rank":2,"votes":134699,"as":"claude-opus-4-6-search","effort":""},"frontierMath":{"v":0.6596491228070176,"as":"Claude Opus 4.6 (max)","effort":"max","date":"2026-06-11","stderr":0.028116467881852295},"aime":{"v":0.9444444444444444,"as":"Claude Opus 4.6 (64k thinking)","effort":"default","date":"2026-02-06","stderr":0.028247294597471075},"gpqa":{"v":0.9053030303030303,"as":"Claude Opus 4.6 (32k thinking)","effort":"default","date":"2026-02-06","stderr":0.016916245311450143},"simpleqa":{"v":0.47,"as":"Claude Opus 4.6 (max)","effort":"max","date":"2026-08-27","stderr":0.015790799515836722},"swebench":{"v":0.7871900826446281,"as":"Claude Opus 4.6 (no thinking)","effort":"none","date":"2026-02-18","stderr":0.01862353730305288},"swebenchAgent":{"v":0.7559999999999999,"as":"Claude 4.6 Opus","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"eqWriting":{"v":1804.1,"as":"claude-opus-4-6","version":"Creative Writing v3"},"eqChat":{"v":1222.7,"as":"claude-opus-4-6","version":"EQ-Bench 4","ciLow":1214.1,"ciHigh":1231.6,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":77.7,"as":"claude-opus-4-6","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"regions":["eu","global"],"n":6,"tps":35,"ttft":2022,"via":"Claude Platform on AWS","reqs":5670},"open":false,"rankable":true,"scored":15},{"id":"anthropic/claude-sonnet-5","name":"Anthropic: Claude Sonnet 5","org":"Anthropic","slug":"claude-sonnet-5","matchKey":"claudesonnet5","blurb":"Sonnet 5 is Anthropic's most capable Sonnet-class model, with frontier performance across coding, agents, and professional work.","created":1782843083,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":2,"out":10,"webSearch":0.01,"alt":{"batch":{"in":1,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":100,"tps":52,"ttft":2503,"reqs":35324},{"name":"Azure","region":"global","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":99.8,"noSchema":1},{"name":"Google","region":"global","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":100,"noSchema":1},{"name":"Amazon Bedrock","region":"global","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":99.9,"noSchema":1},{"name":"Anthropic","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":100,"tps":62,"ttft":1430,"reqs":10747},{"name":"Amazon Bedrock","region":"eu","in":2.2,"out":11,"cacheRead":0.22,"cacheWrite":2.75,"uptime":99.9,"noSchema":1},{"name":"Azure","region":"us","in":2.2,"out":11,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100,"noSchema":1,"tps":64,"ttft":1485,"reqs":75},{"name":"Google","region":"us","in":2.2,"out":11,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100,"noSchema":1},{"name":"Google","region":"eu","in":2.2,"out":11,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100,"noSchema":1,"tps":57,"ttft":2080,"reqs":213},{"name":"Amazon Bedrock","region":"us","in":2.2,"out":11,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100,"noSchema":1,"tps":71,"ttft":2165,"reqs":44}],"scores":{"arenaChat":{"v":1461.2069551775312,"ci":4.6,"rank":51,"votes":35301,"as":"claude-sonnet-5-high","effort":"high"},"arenaWebdev":{"v":1536.7226173776958,"ci":7.2,"rank":33,"votes":9296,"as":"claude-sonnet-5-high","effort":"high"},"arenaVision":{"v":1266.3046736121034,"ci":7.6,"rank":33,"votes":10289,"as":"claude-sonnet-5-high","effort":"high"},"arenaDocument":{"v":1466.136518831971,"ci":7.6,"rank":17,"votes":7411,"as":"claude-sonnet-5-high","effort":"high"},"arenaImageCode":{"v":1542.4941233772238,"ci":9,"rank":17,"votes":5572,"as":"claude-sonnet-5-high","effort":"high"},"arenaSearch":{"v":1194.0719510316353,"ci":6.9,"rank":17,"votes":40230,"as":"claude-sonnet-5-search","effort":""},"frontierMath":{"v":0.656140350877193,"as":"Claude Sonnet 5 (max)","effort":"max","date":"2026-06-30","stderr":0.02818576398907258},"aime":{"v":0.9472222222222222,"as":"Claude Sonnet 5 (xhigh)","effort":"xhigh","date":"2026-07-01","stderr":0.022644349282618412},"gpqa":{"v":0.9053030303030303,"as":"Claude Sonnet 5 (xhigh)","effort":"xhigh","date":"2026-07-01","stderr":0.01783832131341021},"simpleqa":{"v":0.337,"as":"Claude Sonnet 5 (max)","effort":"max","date":"2026-08-27","stderr":0.014955087918653461},"terminal":{"v":0.1242,"as":"Sonnet 5","agent":"Claude Code","effort":"max","ci":0.030600000000000002,"trials":330,"cost":9603.86,"secs":6510.3,"date":"2026-06-30"},"eqWriting":{"v":1790.5,"as":"claude-sonnet-5","version":"Creative Writing v3"},"eqChat":{"v":1236,"as":"claude-sonnet-5","version":"EQ-Bench 4","ciLow":1225.1,"ciHigh":1247.7,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":78.3,"as":"claude-sonnet-5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.2,"regions":["eu","global","us"],"n":10,"tps":52,"ttft":2503,"tpsBest":71,"via":"Claude Platform on AWS","reqs":35324},"open":false,"rankable":true,"scored":14},{"id":"anthropic/claude-opus-4.7","name":"Anthropic: Claude Opus 4.7","org":"Anthropic","slug":"claude-opus-4.7","matchKey":"claudeopus47","blurb":"Opus 4.7 is the next generation of Anthropic's Opus family, built for long-running, asynchronous agents.","created":1776351100,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":25,"webSearch":0.01,"alt":{"batch":{"in":2.5,"out":12.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"tps":45,"ttft":1444,"reqs":2405},{"name":"Azure","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"noSchema":1,"tps":30,"ttft":972,"reqs":1},{"name":"Google","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"noSchema":1},{"name":"Amazon Bedrock","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"noSchema":1},{"name":"Anthropic","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"tps":48,"ttft":1234,"reqs":630},{"name":"Google","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"noSchema":1},{"name":"Amazon Bedrock","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"noSchema":1,"tps":52,"ttft":2007,"reqs":375},{"name":"Google","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"noSchema":1,"tps":80,"ttft":2362,"reqs":3193}],"scores":{"arenaChat":{"v":1501.7492535919955,"ci":3.9,"rank":3,"votes":60002,"as":"claude-opus-4-7-high","effort":"high"},"arenaWebdev":{"v":1557.3721202968718,"ci":6,"rank":25,"votes":15914,"as":"claude-opus-4-7","effort":""},"arenaVision":{"v":1300.9352659167705,"ci":6.8,"rank":3,"votes":21092,"as":"claude-opus-4-7-high","effort":"high"},"arenaDocument":{"v":1495.0390463625913,"ci":6.4,"rank":6,"votes":22192,"as":"claude-opus-4-7","effort":""},"arenaImageCode":{"v":1575.9295067235266,"ci":9.4,"rank":13,"votes":6401,"as":"claude-opus-4-7-high","effort":"high"},"arenaSearch":{"v":1233.2545271621282,"ci":5.4,"rank":4,"votes":91394,"as":"claude-opus-4-7","effort":""},"frontierMath":{"v":0.7017543859649122,"as":"Claude Opus 4.7 (max)","effort":"max","date":"2026-06-10","stderr":0.027146911721220194},"aime":{"v":0.978,"as":"Claude Opus 4.7 (xhigh)","effort":"xhigh","date":"2026-04-17","stderr":0.022},"gpqa":{"v":0.9015151515151515,"as":"Claude Opus 4.7 (xhigh)","effort":"xhigh","date":"2026-04-17","stderr":0.018464275887293338},"simpleqa":{"v":0.517,"as":"Claude Opus 4.7 (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015810153729833274},"swebench":{"v":0.8347107438016529,"as":"Claude Opus 4.7 (max)","effort":"max","date":"2026-04-20","stderr":0.016901169179529604},"eqWriting":{"v":1907.1,"as":"claude-opus-4-7","version":"Creative Writing v3"},"eqChat":{"v":1311.3,"as":"claude-opus-4-7","version":"EQ-Bench 4","ciLow":1299.3,"ciHigh":1321.8,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":81.8,"as":"claude-opus-4-7","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"regions":["eu","global","us"],"n":8,"tps":80,"ttft":2362,"via":"Google","reqs":3193},"open":false,"rankable":true,"scored":14},{"id":"openai/gpt-5.4","name":"OpenAI: GPT-5.4","org":"OpenAI","slug":"gpt-5.4","matchKey":"gpt54","blurb":"GPT-5.4 is OpenAI’s latest frontier model, unifying the Codex and GPT lines into a single system.","created":1772734352,"cutoff":null,"hf":null,"ctx":1050000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":2.5,"out":15,"webSearch":0.01,"alt":{"batch":{"in":1.25,"out":7.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":1.25,"out":7.5,"cacheRead":0.125,"uptime":89.6},{"name":"Azure","in":2.5,"out":15,"cacheRead":0.25,"uptime":100},{"name":"OpenAI","in":2.5,"out":15,"cacheRead":0.25,"uptime":100},{"name":"Azure","region":"us","in":2.75,"out":16.5,"cacheRead":0.275},{"name":"Azure","region":"eu","in":2.75,"out":16.5,"cacheRead":0.275,"uptime":100,"tps":64,"ttft":2285,"reqs":1075},{"name":"Amazon Bedrock","region":"us","in":2.75,"out":16.5,"cacheRead":0.275,"noSchema":1,"tps":6,"ttft":1605,"reqs":61},{"name":"OpenAI","in":5,"out":30,"cacheRead":0.5,"uptime":100,"tps":31,"ttft":1048,"reqs":8700}],"scores":{"arenaChat":{"v":1476.3897159979338,"ci":3.9,"rank":26,"votes":60537,"as":"gpt-5.4-high","effort":"high"},"arenaWebdev":{"v":1462.9864068566735,"ci":19.4,"rank":54,"votes":1277,"as":"gpt-5.4-high (codex-harness)","effort":"high"},"arenaVision":{"v":1284.7878972861884,"ci":6.3,"rank":15,"votes":25357,"as":"gpt-5.4-high","effort":"high"},"arenaDocument":{"v":1471.0043734130195,"ci":6,"rank":14,"votes":33331,"as":"gpt-5.4","effort":""},"arenaImageCode":{"v":1445.4663065618938,"ci":13.4,"rank":38,"votes":2433,"as":"gpt-5.4","effort":""},"arenaSearch":{"v":1197.1574686457088,"ci":5.4,"rank":16,"votes":110116,"as":"gpt-5.4-search","effort":""},"frontierMath":{"v":0.7859649122807018,"as":"GPT-5.4 (xhigh)","effort":"xhigh","date":"2026-06-11","stderr":0.024338000553262393},"aime":{"v":0.9777777777777777,"as":"GPT-5.4 (high)","effort":"high","date":"2026-07-15","stderr":0.022222222222222223},"gpqa":{"v":0.933,"as":"GPT-5.4 (xhigh)","effort":"xhigh","date":"2026-03-06","stderr":0.018},"simpleqa":{"v":0.451,"as":"GPT-5.4 (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015743152379585443},"swebench":{"v":0.768595041322314,"as":"GPT-5.4 (high)","effort":"high","date":"2026-03-06","stderr":0.019189407003691976},"eqWriting":{"v":1835.6,"as":"gpt-5.4","version":"Creative Writing v3"},"eqChat":{"v":1271.8,"as":"openai/gpt-5.4","version":"EQ-Bench 4","ciLow":1261.5,"ciHigh":1281.4,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":78.3,"as":"gpt-5.4","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.125,"regions":["eu","us"],"n":7,"tps":31,"ttft":1048,"tpsBest":64,"via":"OpenAI","reqs":8700},"open":false,"rankable":true,"scored":14},{"id":"openai/gpt-5.5","name":"OpenAI: GPT-5.5","org":"OpenAI","slug":"gpt-5.5","matchKey":"gpt55","blurb":"GPT-5.5 is OpenAI’s frontier model designed for complex professional workloads, building on GPT-5.4 with stronger reasoning, higher reliability, and improved token efficiency on hard tasks.","created":1777051893,"cutoff":"2025-12-01","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":5,"out":30,"webSearch":0.01,"alt":{"batch":{"in":2.5,"out":15}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":2.5,"out":15,"cacheRead":0.25,"uptime":99.8},{"name":"Azure","in":5,"out":30,"cacheRead":0.5,"uptime":100},{"name":"OpenAI","in":5,"out":30,"cacheRead":0.5,"uptime":99.9},{"name":"Azure","region":"us","in":5.5,"out":33,"cacheRead":0.55,"uptime":100},{"name":"Azure","region":"eu","in":5.5,"out":33,"cacheRead":0.55,"uptime":100,"tps":55,"ttft":2681,"reqs":3},{"name":"Amazon Bedrock","region":"us","in":5.5,"out":33,"cacheRead":0.55,"noSchema":1},{"name":"OpenAI","in":12.5,"out":75,"cacheRead":1.25,"uptime":100,"tps":63,"ttft":1758,"reqs":5281}],"scores":{"arenaChat":{"v":1481.756564434564,"ci":3.8,"rank":20,"votes":64924,"as":"gpt-5.5-high","effort":"high"},"arenaWebdev":{"v":1510.1450492200738,"ci":6.1,"rank":41,"votes":14946,"as":"gpt-5.5-xhigh (codex-harness)","effort":"xhigh"},"arenaVision":{"v":1287.1265092166132,"ci":6.3,"rank":13,"votes":23423,"as":"gpt-5.5","effort":""},"arenaDocument":{"v":1485.7193136054452,"ci":6.2,"rank":8,"votes":21279,"as":"gpt-5.5","effort":""},"arenaImageCode":{"v":1532.3635000424065,"ci":9.4,"rank":22,"votes":6375,"as":"gpt-5.5-xhigh (codex-harness)","effort":"xhigh"},"arenaSearch":{"v":1242.052705813796,"ci":5.3,"rank":3,"votes":89873,"as":"gpt-5.5-search","effort":""},"frontierMath":{"v":0.8526315789473684,"as":"GPT-5.5 (xhigh)","effort":"xhigh","date":"2026-06-11","stderr":0.021034091168852485},"aime":{"v":1,"as":"GPT-5.5 (xhigh)","effort":"xhigh","date":"2026-04-24","stderr":0},"gpqa":{"v":0.9400252525252525,"as":"GPT-5.5 (xhigh)","effort":"xhigh","date":"2026-04-24","stderr":0.015469771177277509},"simpleqa":{"v":0.63,"as":"GPT-5.5 (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015275252316519465},"swebench":{"v":0.8057851239669421,"as":"GPT-5.5 (xhigh)","effort":"xhigh","date":"2026-04-24","stderr":0.018000198869416815},"eqWriting":{"v":1843.5,"as":"gpt-5.5","version":"Creative Writing v3"},"eqChat":{"v":1314.8,"as":"openai/gpt-5.5","version":"EQ-Bench 4","ciLow":1301.7,"ciHigh":1325.4,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":78.2,"as":"gpt-5.5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.25,"regions":["eu","us"],"n":7,"tps":63,"ttft":1758,"tpsBest":67,"via":"OpenAI","reqs":5281},"open":false,"rankable":true,"scored":14},{"id":"openai/gpt-5.6-sol","name":"OpenAI: GPT-5.6 Sol","org":"OpenAI","slug":"gpt-5.6-sol","matchKey":"gpt56sol","blurb":"GPT-5.6 Sol is the flagship model in OpenAI's GPT-5.6 series.","created":1783590850,"cutoff":"2026-02-16","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":2,"out":10,"webSearch":0.01,"alt":{"batch":{"in":1,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":1,"out":5,"cacheRead":0.1,"cacheWrite":1.25,"uptime":87.4},{"name":"OpenAI","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":99.8},{"name":"OpenAI","in":4,"out":20,"cacheRead":0.4,"cacheWrite":5,"uptime":100,"tps":49,"ttft":3658,"reqs":65772},{"name":"Amazon Bedrock","region":"us","in":4.4,"out":22,"cacheRead":0.44,"cacheWrite":5.5,"uptime":99.9,"noSchema":1,"tps":75,"ttft":1347,"reqs":2069},{"name":"Azure","in":5,"out":30,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100},{"name":"Azure","region":"us","in":5.5,"out":33,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100},{"name":"Azure","region":"eu","in":5.5,"out":33,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"tps":17,"ttft":19041,"reqs":51}],"scores":{"arenaChat":{"v":1483.474347288669,"ci":4.9,"rank":18,"votes":27069,"as":"gpt-5.6-sol-xhigh","effort":"xhigh"},"arenaWebdev":{"v":1617.3213915191732,"ci":7,"rank":14,"votes":11916,"as":"gpt-5.6-sol-xhigh (codex-harness)","effort":"xhigh"},"arenaVision":{"v":1285.9461347353815,"ci":8.4,"rank":14,"votes":7729,"as":"gpt-5.6-sol-xhigh","effort":"xhigh"},"arenaDocument":{"v":1482.7123255544448,"ci":8.8,"rank":10,"votes":4818,"as":"gpt-5.6-sol-xhigh","effort":"xhigh"},"arenaImageCode":{"v":1603.5062307649641,"ci":12.6,"rank":8,"votes":2515,"as":"gpt-5.6-sol-xhigh (codex-harness)","effort":"xhigh"},"arenaSearch":{"v":1257.2603500980206,"ci":7.5,"rank":1,"votes":29663,"as":"gpt-5.6-sol-xhigh","effort":"xhigh"},"frontierMath":{"v":0.8912280701754386,"as":"GPT-5.6 Sol (max)","effort":"max","date":"2026-07-09","stderr":0.018475392571140503},"aime":{"v":1,"as":"GPT-5.6 Sol (max)","effort":"max","date":"2026-07-09","stderr":0},"gpqa":{"v":0.9349747474747475,"as":"GPT-5.6 Sol (max)","effort":"max","date":"2026-07-09","stderr":0.01570109070867593},"simpleqa":{"v":0.697,"as":"GPT-5.6 Sol (max)","effort":"max","date":"2026-08-10","stderr":0.014539683710535193},"terminal":{"v":0.37270000000000003,"as":"GPT-5.6 Sol","agent":"Codex","effort":"max","ci":0.0378,"trials":330,"cost":2541.7,"secs":2388.4,"date":"2026-06-26"},"eqWriting":{"v":1963.4,"as":"gpt-5.6-sol","version":"Creative Writing v3"},"eqChat":{"v":1250.3,"as":"openai/gpt-5.6-sol","version":"EQ-Bench 4","ciLow":1239.6,"ciHigh":1263.2,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":81.7,"as":"gpt-5.6-sol","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.1,"regions":["eu","us"],"n":7,"tps":49,"ttft":3658,"tpsBest":75,"via":"OpenAI","reqs":65772},"open":false,"rankable":true,"scored":14},{"id":"anthropic/claude-fable-5","name":"Anthropic: Claude Fable 5","org":"Anthropic","slug":"claude-fable-5","matchKey":"claudefable5","blurb":"Claude Fable 5 is a Mythos-class model from Anthropic, built for autonomous knowledge work and coding.","created":1781007515,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":10,"out":50,"webSearch":0.01,"alt":{"batch":{"in":5,"out":25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":99.7,"tps":53,"ttft":6729,"reqs":39},{"name":"Azure","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":84,"noSchema":1,"tps":29,"ttft":5620,"reqs":25},{"name":"Amazon Bedrock","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"noSchema":1,"tps":2,"ttft":2253,"reqs":2},{"name":"Google","region":"global","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":100,"noSchema":1},{"name":"Anthropic","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":99.3,"tps":51,"ttft":4604,"reqs":1691},{"name":"Google","region":"eu","in":11,"out":55,"cacheRead":1.1,"cacheWrite":13.75,"uptime":100,"noSchema":1,"tps":16,"ttft":2282,"reqs":4}],"scores":{"arenaChat":{"v":1505.6827180827381,"ci":4.8,"rank":1,"votes":30057,"as":"claude-fable-5-high","effort":"high"},"arenaWebdev":{"v":1627.9838820923837,"ci":7.5,"rank":10,"votes":10081,"as":"claude-fable-5-high","effort":"high"},"arenaVision":{"v":1309.503174212711,"ci":7.9,"rank":1,"votes":11304,"as":"claude-fable-5-high","effort":"high"},"arenaDocument":{"v":1496.0822362465265,"ci":7.9,"rank":5,"votes":8497,"as":"claude-fable-5-high","effort":"high"},"arenaImageCode":{"v":1622.8150714125632,"ci":11,"rank":6,"votes":3806,"as":"claude-fable-5-high","effort":"high"},"arenaSearch":{"v":1230.2225243963933,"ci":7.9,"rank":5,"votes":41795,"as":"claude-fable-5-high","effort":"high"},"frontierMath":{"v":0.8701754385964913,"as":"Claude Fable 5 (max)","effort":"max","date":"2026-06-09","stderr":0.019944477920133024},"aime":{"v":1,"as":"Claude Fable 5 (high)","effort":"high","date":"2026-08-06","stderr":0},"gpqa":{"v":0.8585858585858586,"as":"Claude Fable 5 (max)","effort":"max","date":"2026-08-06","stderr":0.024825909793343353},"simpleqa":{"v":0.707,"as":"Claude Fable 5 (xhigh)","effort":"xhigh","date":"2026-08-10","stderr":0.014399942998441292},"terminal":{"v":0.44549999999999995,"as":"Fable 5","agent":"Claude Code","effort":"max","ci":0.0385,"trials":330,"cost":7265.01,"secs":4202.8,"date":"2026-06-09"},"eqWriting":{"v":1934.6,"as":"claude-fable-5","version":"Creative Writing v3"},"eqChat":{"v":1340.4,"as":"claude-fable-5","version":"EQ-Bench 4","ciLow":1328.5,"ciHigh":1355.2,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":83,"as":"claude-fable-5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":1,"regions":["eu","global"],"n":6,"tps":51,"ttft":4604,"tpsBest":53,"via":"Anthropic","reqs":1691},"open":false,"rankable":true,"scored":14},{"id":"google/gemini-3.1-pro-preview","name":"Google: Gemini 3.1 Pro Preview","org":"Google","slug":"gemini-3.1-pro-preview","matchKey":"gemini31propreview","blurb":"Gemini 3.1 Pro Preview is Google’s frontier reasoning model, delivering enhanced software engineering performance, improved agentic reliability, and more efficient token usage across complex workflows.","created":1771509627,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["audio","file","image","text","video"],"out":["text"],"price":{"in":2,"out":12,"imageIn":0.000002,"audioIn":2,"webSearch":0.014,"alt":{"batch":{"in":1,"out":6}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Google","in":1,"out":6,"cacheRead":0.1,"cacheWrite":0.1875,"uptime":6.7},{"name":"Google AI Studio","in":1,"out":6,"cacheRead":0.1,"cacheWrite":0.1875,"uptime":99.9},{"name":"Google","region":"global","in":2,"out":12,"cacheRead":0.2,"cacheWrite":0.375,"uptime":96.4},{"name":"Google AI Studio","in":2,"out":12,"cacheRead":0.2,"cacheWrite":0.375,"uptime":99.9},{"name":"Google","in":3.6,"out":21.6,"cacheRead":0.36,"cacheWrite":0.675,"uptime":97.3,"tps":95,"ttft":2867,"reqs":16184,"region":"global"},{"name":"Google AI Studio","in":3.6,"out":21.6,"cacheRead":0.36,"cacheWrite":0.675,"tps":100,"ttft":3021,"reqs":1785}],"scores":{"arenaChat":{"v":1486.809321925295,"ci":3.2,"rank":15,"votes":106951,"as":"gemini-3.1-pro-preview","effort":""},"arenaWebdev":{"v":1446.6380535473631,"ci":5.2,"rank":57,"votes":22891,"as":"gemini-3.1-pro-preview","effort":""},"arenaVision":{"v":1278.693223560465,"ci":5.5,"rank":26,"votes":40691,"as":"gemini-3.1-pro-preview","effort":""},"arenaDocument":{"v":1443.9187054358893,"ci":5.2,"rank":31,"votes":49457,"as":"gemini-3.1-pro-preview","effort":""},"arenaImageCode":{"v":1478.3065001933755,"ci":8.9,"rank":32,"votes":7615,"as":"gemini-3.1-pro-preview","effort":""},"arenaSearch":{"v":1210.4636662372704,"ci":5.1,"rank":9,"votes":113282,"as":"gemini-3.1-pro-grounding","effort":""},"frontierMath":{"v":0.5964912280701754,"as":"Gemini 3.1 Pro Preview","effort":"","date":"2026-06-11","stderr":0.02911181956525014},"aime":{"v":0.956,"as":"Gemini 3.1 Pro Preview","effort":"","date":"2026-02-20","stderr":0.031},"gpqa":{"v":0.9444444444444444,"as":"Gemini 3.1 Pro Preview","effort":"","date":"2026-08-06","stderr":0.016319950700767385},"simpleqa":{"v":0.735,"as":"Gemini 3.1 Pro Preview","effort":"","date":"2026-08-10","stderr":0.013963164754810065},"swebench":{"v":0.756198347107438,"as":"Gemini 3.1 Pro Preview","effort":"","date":"2026-02-24","stderr":0.019537213360116365},"eqWriting":{"v":1488.9,"as":"gemini-3.1-pro-preview","version":"Creative Writing v3"},"eqChat":{"v":1142.3,"as":"google/gemini-3.1-pro-preview","version":"EQ-Bench 4","ciLow":1130,"ciHigh":1153.9,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":68.2,"as":"gemini-3.1-pro-preview","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.1,"regions":["global"],"n":6,"tps":95,"ttft":2867,"tpsBest":100,"via":"Google","reqs":16184},"open":false,"rankable":true,"scored":14},{"id":"openai/gpt-5.2","name":"OpenAI: GPT-5.2","org":"OpenAI","slug":"gpt-5.2","matchKey":"gpt52","blurb":"GPT-5.2 is the latest frontier-grade model in the GPT-5 series, offering stronger agentic and long context perfomance compared to GPT-5.1.","created":1765389775,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":1.75,"out":14,"webSearch":0.01,"alt":{"batch":{"in":0.875,"out":7}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.875,"out":7,"cacheRead":0.0875,"uptime":100},{"name":"Azure","in":1.75,"out":14,"cacheRead":0.175,"uptime":100,"tps":32,"ttft":3207,"reqs":647},{"name":"OpenAI","in":1.75,"out":14,"cacheRead":0.175,"uptime":99.7},{"name":"OpenAI","in":3.5,"out":28,"cacheRead":0.35,"uptime":100,"tps":31,"ttft":2226,"reqs":10918}],"scores":{"arenaChat":{"v":1437.582781224327,"ci":3.6,"rank":89,"votes":47538,"as":"gpt-5.2-high","effort":"high"},"arenaWebdev":{"v":1416.642829921493,"ci":22.9,"rank":70,"votes":1055,"as":"gpt-5.2","effort":""},"arenaVision":{"v":1244.0403441238918,"ci":6.3,"rank":50,"votes":17926,"as":"gpt-5.2-high","effort":"high"},"arenaDocument":{"v":1404.5908350822938,"ci":9.5,"rank":41,"votes":7108,"as":"gpt-5.2-high","effort":"high"},"arenaSearch":{"v":1206.9460713540655,"ci":5.6,"rank":11,"votes":52712,"as":"gpt-5.2-search","effort":""},"frontierMath":{"v":0.674,"as":"GPT-5.2 (xhigh)","effort":"xhigh","date":"2026-06-11","stderr":0.028},"aime":{"v":0.9611111111111111,"as":"GPT-5.2 (high)","effort":"high","date":"2025-12-11","stderr":0.025997226390952678},"gpqa":{"v":0.914,"as":"GPT-5.2 (xhigh)","effort":"xhigh","date":"2025-12-13","stderr":0.018},"simpleqa":{"v":0.371,"as":"GPT-5.2 (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015283736211823095},"swebench":{"v":0.737603305785124,"as":"GPT-5.2 (high)","effort":"high","date":"2026-02-12","stderr":0.020017832064928277},"swebenchAgent":{"v":0.728,"as":"GPT 5.2","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"bfcl":{"v":0.5587,"as":"GPT-5.2-2025-12-11 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1699.8,"as":"gpt-5.2","version":"Creative Writing v3"},"eqLongform":{"v":71.6,"as":"gpt-5.2","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0875,"n":4,"tps":31,"ttft":2226,"tpsBest":32,"via":"OpenAI","reqs":10918},"open":false,"rankable":true,"scored":14},{"id":"moonshotai/kimi-k2.6","name":"MoonshotAI: Kimi K2.6","org":"MoonshotAI","slug":"kimi-k2.6","matchKey":"kimik26","blurb":"Kimi K2.6 is Moonshot AI's next-generation multimodal model, designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration.","created":1776699402,"cutoff":null,"hf":"moonshotai/Kimi-K2.6","ctx":262144,"maxOut":235929,"in":["text","image"],"out":["text"],"price":{"in":0.95,"out":4},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Chutes","quant":"int4","maxOut":65535,"in":0.5,"out":2.85,"cacheRead":0.05,"uptime":98.1,"tps":24,"ttft":2494,"reqs":2558},{"name":"Inceptron","quant":"int4","in":0.5181,"out":2.97,"cacheRead":0.0999,"uptime":99.3,"tps":28,"ttft":708,"reqs":2327,"region":"eu"},{"name":"DigitalOcean","in":0.57,"out":2.4,"cacheRead":0.114,"uptime":99.5,"tps":7,"ttft":638,"reqs":1650},{"name":"Decart","quant":"fp4","in":0.5865,"out":2.4696,"cacheRead":0.0988,"uptime":96.2,"tps":83,"ttft":608,"reqs":648},{"name":"StreamLake","quant":"fp8","ctx":256000,"maxOut":230400,"in":0.5985,"out":2.52,"cacheRead":0.1008,"uptime":97.9,"tps":42,"ttft":1183,"reqs":2327},{"name":"CoreWeave","quant":"fp4","in":0.65,"out":3.41,"cacheRead":0.15,"uptime":99.7,"tps":62,"ttft":345,"reqs":2524},{"name":"Crusoe","quant":"bf16","in":0.7,"out":3.5,"cacheRead":0.35,"uptime":98.2,"tps":18,"ttft":341,"reqs":972},{"name":"DeepInfra","quant":"fp4","maxOut":16384,"in":0.75,"out":3.5,"cacheRead":0.15,"uptime":97.4,"tps":30,"ttft":1218,"reqs":245},{"name":"Venice","quant":"int4","ctx":256000,"maxOut":65536,"in":0.75,"out":3.5,"cacheRead":0.16,"uptime":89,"tps":21,"ttft":1603,"reqs":414},{"name":"Parasail","quant":"int4","in":0.75,"out":3.5,"cacheRead":0.16,"uptime":99.7,"tps":62,"ttft":485,"reqs":1488},{"name":"SiliconFlow","quant":"fp8","in":0.77,"out":3.4,"cacheRead":0.14,"uptime":99.8,"tps":27,"ttft":781,"reqs":1464},{"name":"Novita","in":0.8,"out":3.4,"cacheRead":0.16,"uptime":99.6,"tps":40,"ttft":1470,"reqs":2066},{"name":"GMICloud","quant":"fp8","in":0.855,"out":3.6,"cacheRead":0.144,"uptime":87.1,"tps":36,"ttft":2342,"reqs":664},{"name":"Baidu","quant":"fp4","in":0.95,"out":4,"cacheRead":0.16,"uptime":100,"tps":61,"ttft":690,"reqs":1295},{"name":"AtlasCloud","quant":"int4","in":0.95,"out":4,"cacheRead":0.16,"uptime":98.1,"tps":52,"ttft":1032,"reqs":709},{"name":"Cloudflare","in":0.95,"out":4,"cacheRead":0.16,"uptime":99.9,"tps":54,"ttft":450,"reqs":763},{"name":"Moonshot AI","quant":"int4","in":0.95,"out":4,"cacheRead":0.16,"uptime":99.9,"tps":37,"ttft":1076,"reqs":5289},{"name":"Sail Research","quant":"int4","in":1,"out":4,"cacheRead":0.2,"uptime":100,"tps":33,"ttft":424,"reqs":403},{"name":"Phala","in":1.09,"out":4.6,"cacheRead":0.37,"uptime":99.3,"tps":42,"ttft":1479,"reqs":1264},{"name":"BaseTen","quant":"fp4","ctx":262000,"maxOut":235800,"in":0.95,"out":4,"cacheRead":0.16,"uptime":97.9},{"name":"Fireworks","in":0.95,"out":4,"cacheRead":0.16,"uptime":0,"tps":53,"ttft":679,"reqs":80},{"name":"BaseTen","quant":"fp4","ctx":262000,"maxOut":235800,"in":0.95,"out":4,"cacheRead":0.16,"tps":25,"ttft":537,"reqs":9,"region":"us"}],"scores":{"arenaChat":{"v":1460.4445059653572,"ci":4.5,"rank":52,"votes":37502,"as":"kimi-k2.6","effort":""},"arenaWebdev":{"v":1508.7735440931424,"ci":7.4,"rank":42,"votes":9585,"as":"kimi-k2.6","effort":""},"arenaVision":{"v":1262.369449782495,"ci":7.1,"rank":37,"votes":15347,"as":"kimi-k2.6","effort":""},"arenaDocument":{"v":1450.7397099818368,"ci":7.6,"rank":27,"votes":11291,"as":"kimi-k2.6","effort":""},"arenaImageCode":{"v":1521.3981271365594,"ci":11.7,"rank":24,"votes":3355,"as":"kimi-k2.6","effort":""},"frontierMath":{"v":0.5719298245614035,"as":"Kimi K2.6","effort":"","date":"2026-06-10","stderr":0.029360921879029937},"aime":{"v":0.9611111111111111,"as":"Kimi K2.6","effort":"","date":"2026-05-02","stderr":0.023777636174410793},"gpqa":{"v":0.9078282828282829,"as":"Kimi K2.6","effort":"","date":"2026-05-01","stderr":0.0171745162866666},"simpleqa":{"v":0.349,"as":"Kimi K2.6","effort":"","date":"2026-08-10","stderr":0.015080663991563052},"swebench":{"v":0.7665289256198347,"as":"Kimi K2.6","effort":"","date":"2026-05-08","stderr":0.01924895909142592},"eqWriting":{"v":1721.3,"as":"moonshotai/Kimi-K2.6","version":"Creative Writing v3"},"eqChat":{"v":1201.6,"as":"moonshotai/kimi-k2.6","version":"EQ-Bench 4","ciLow":1192.8,"ciHigh":1210.1,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":78.5,"as":"moonshotai/Kimi-K2.6","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.05,"n":22,"tps":37,"ttft":1076,"tpsBest":103,"via":"Moonshot AI","reqs":5289,"regions":["eu","us"]},"open":true,"rankable":true,"scored":13},{"id":"anthropic/claude-opus-4.8","name":"Anthropic: Claude Opus 4.8","org":"Anthropic","slug":"claude-opus-4.8","matchKey":"claudeopus48","blurb":"Claude Opus 4.8 is Anthropic's most capable generally available model in the Opus family.","created":1779905091,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":25,"webSearch":0.01,"alt":{"batch":{"in":2.5,"out":12.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"Claude Platform on AWS","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"tps":57,"ttft":1537,"reqs":3190},{"name":"Azure","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":96.9,"noSchema":1},{"name":"Amazon Bedrock","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":98.7,"noSchema":1},{"name":"Google","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9,"noSchema":1},{"name":"Anthropic","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9},{"name":"Azure","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"noSchema":1,"tps":30,"ttft":1507,"reqs":1},{"name":"Amazon Bedrock","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"noSchema":1},{"name":"Google","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"noSchema":1},{"name":"Amazon Bedrock","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":99.9,"noSchema":1,"tps":67,"ttft":2500,"reqs":441},{"name":"Google","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"noSchema":1,"tps":64,"ttft":3651,"reqs":1031},{"name":"Anthropic","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":100,"tps":47,"ttft":1599,"reqs":1089}],"scores":{"arenaChat":{"v":1481.3781065574426,"ci":4.2,"rank":21,"votes":52535,"as":"claude-opus-4-8-high","effort":"high"},"arenaWebdev":{"v":1558.8327870970375,"ci":6.5,"rank":24,"votes":13787,"as":"claude-opus-4-8-high","effort":"high"},"arenaVision":{"v":1283.4203054080024,"ci":7,"rank":18,"votes":15274,"as":"claude-opus-4-8-high","effort":"high"},"arenaDocument":{"v":1474.788449846873,"ci":6.9,"rank":12,"votes":11551,"as":"claude-opus-4-8-high","effort":"high"},"arenaSearch":{"v":1204.3009106749619,"ci":6.4,"rank":12,"votes":70998,"as":"claude-opus-4-8","effort":""},"frontierMath":{"v":0.8,"as":"Claude Opus 4.8","effort":"","date":"2026-06-10","stderr":0.023735633163877067},"aime":{"v":0.9833333333333333,"as":"Claude Opus 4.8","effort":"","date":"2026-06-07","stderr":0.014101901870444154},"gpqa":{"v":0.9103535353535354,"as":"Claude Opus 4.8","effort":"","date":"2026-06-07","stderr":0.019157125794612026},"simpleqa":{"v":0.53,"as":"Claude Opus 4.8","effort":"","date":"2026-08-27","stderr":0.015790799515836722},"terminal":{"v":0.2364,"as":"Opus 4.8","agent":"Claude Code","effort":"max","ci":0.0356,"trials":330,"cost":6481.26,"secs":5163,"date":"2026-05-28"},"eqWriting":{"v":1835.2,"as":"claude-opus-4-8","version":"Creative Writing v3"},"eqChat":{"v":1281.4,"as":"claude-opus-4-8","version":"EQ-Bench 4","ciLow":1270.3,"ciHigh":1292.5,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":80.8,"as":"claude-opus-4-8","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"regions":["eu","global","us"],"n":11,"tps":57,"ttft":1537,"tpsBest":67,"via":"Claude Platform on AWS","reqs":3190},"open":false,"rankable":true,"scored":13},{"id":"anthropic/claude-opus-5","name":"Anthropic: Claude Opus 5","org":"Anthropic","slug":"claude-opus-5","matchKey":"claudeopus5","blurb":"Claude Opus 5 is Anthropic’s flagship model for demanding reasoning, coding, and long-horizon agentic work.","created":1784912544,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":25,"webSearch":0.01,"alt":{"batch":{"in":2.5,"out":12.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9,"tps":52,"ttft":4865,"reqs":15234},{"name":"Google","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9,"noSchema":1},{"name":"Amazon Bedrock","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9,"noSchema":1},{"name":"Azure","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9,"noSchema":1},{"name":"Anthropic","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9},{"name":"Amazon Bedrock","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"noSchema":1},{"name":"Azure","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":99.9,"noSchema":1,"tps":82,"ttft":3597,"reqs":195},{"name":"Google","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"noSchema":1},{"name":"Google","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":99.9,"noSchema":1,"tps":47,"ttft":4571,"reqs":8},{"name":"Amazon Bedrock","region":"us","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"noSchema":1,"tps":56,"ttft":1096,"reqs":9},{"name":"Anthropic","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":100,"tps":60,"ttft":3822,"reqs":10347}],"scores":{"arenaChat":{"v":1492.8999328031769,"ci":4.2,"rank":10,"votes":42617,"as":"claude-opus-5-high","effort":"high"},"arenaWebdev":{"v":1686.9942451998663,"ci":7.5,"rank":3,"votes":12087,"as":"claude-opus-5-max","effort":"max"},"arenaVision":{"v":1289.0268908903133,"ci":7.6,"rank":10,"votes":11599,"as":"claude-opus-5-high","effort":"high"},"arenaDocument":{"v":1516.2663910026604,"ci":7,"rank":1,"votes":8794,"as":"claude-opus-5-high","effort":"high"},"arenaImageCode":{"v":1665.1465139387792,"ci":13.3,"rank":3,"votes":2490,"as":"claude-opus-5-max","effort":"max"},"frontierMath":{"v":0.856140350877193,"as":"Claude Opus 5 (max)","effort":"max","date":"2026-07-24","stderr":0.020824894575343943},"aime":{"v":0.9888888888888889,"as":"Claude Opus 5 (max)","effort":"max","date":"2026-07-24","stderr":0.011111111111111112},"gpqa":{"v":0.9387626262626263,"as":"Claude Opus 5 (max)","effort":"max","date":"2026-07-24","stderr":0.014782015056089346},"simpleqa":{"v":0.599,"as":"Claude Opus 5 (max)","effort":"max","date":"2026-08-10","stderr":0.01550610974549837},"terminal":{"v":0.5394,"as":"Opus 5","agent":"Claude Code","effort":"xhigh","ci":0.0317,"trials":330,"cost":6086.22,"secs":4514.7,"date":"2026-07-24"},"eqWriting":{"v":2120.6,"as":"claude-opus-5","version":"Creative Writing v3"},"eqChat":{"v":1385,"as":"claude-opus-5","version":"EQ-Bench 4","ciLow":1362.4,"ciHigh":1412,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":86.3,"as":"claude-opus-5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"regions":["eu","global","us"],"n":11,"tps":52,"ttft":4865,"tpsBest":82,"via":"Claude Platform on AWS","reqs":15234},"open":false,"rankable":true,"scored":13},{"id":"anthropic/claude-sonnet-4.6","name":"Anthropic: Claude Sonnet 4.6","org":"Anthropic","slug":"claude-sonnet-4.6","matchKey":"claudesonnet46","blurb":"Sonnet 4.6 is Anthropic's most capable Sonnet-class model yet, with frontier performance across coding, agents, and professional work.","created":1771342990,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":3,"out":15,"webSearch":0.01,"alt":{"batch":{"in":1.5,"out":7.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","high","medium","low"]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":100,"tps":40,"ttft":1107,"reqs":30200},{"name":"Azure","region":"global","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":94.1,"noSchema":1},{"name":"Anthropic","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":100,"tps":31,"ttft":1226,"reqs":19017},{"name":"Amazon Bedrock","region":"global","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":99.9},{"name":"Google","region":"global","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":100,"noSchema":1},{"name":"Amazon Bedrock","region":"us","in":3.3,"out":16.5,"cacheRead":0.33,"cacheWrite":4.125,"uptime":100},{"name":"Amazon Bedrock","region":"eu","in":3.3,"out":16.5,"cacheRead":0.33,"cacheWrite":4.125,"uptime":100,"tps":45,"ttft":1151,"reqs":19},{"name":"Google","region":"eu","in":3.3,"out":16.5,"cacheRead":0.33,"cacheWrite":4.125,"uptime":100,"noSchema":1},{"name":"Google","region":"us","in":3.3,"out":16.5,"cacheRead":0.33,"cacheWrite":4.125,"uptime":100,"noSchema":1,"tps":13,"ttft":1271,"reqs":2}],"scores":{"arenaChat":{"v":1472.57427288615,"ci":3.6,"rank":37,"votes":66208,"as":"claude-sonnet-4-6","effort":""},"arenaWebdev":{"v":1520.9868241527552,"ci":5.3,"rank":36,"votes":21642,"as":"claude-sonnet-4-6","effort":""},"arenaVision":{"v":1275.433149395297,"ci":6.4,"rank":29,"votes":25552,"as":"claude-sonnet-4-6","effort":""},"arenaDocument":{"v":1482.23522561877,"ci":5.9,"rank":11,"votes":57813,"as":"claude-sonnet-4-6","effort":""},"arenaImageCode":{"v":1530.0480338940542,"ci":9.1,"rank":23,"votes":7726,"as":"claude-sonnet-4-6","effort":""},"arenaSearch":{"v":1221.231466167688,"ci":4.8,"rank":7,"votes":134905,"as":"claude-sonnet-4-6-search","effort":""},"aime":{"v":0.857936507936508,"as":"Claude Sonnet 4.6 (32k thinking)","effort":"default","date":"2026-02-20","stderr":0.04539927553202047},"gpqa":{"v":0.8737373737373737,"as":"Claude Sonnet 4.6 (32k thinking)","effort":"default","date":"2026-02-20","stderr":0.020034282226331888},"simpleqa":{"v":0.355,"as":"Claude Sonnet 4.6 (high)","effort":"high","date":"2026-08-10","stderr":0.015139491543780595},"swebench":{"v":0.7520661157024794,"as":"Claude Sonnet 4.6 (no thinking)","effort":"none","date":"2026-02-21","stderr":0.01964818270139032},"eqWriting":{"v":1804.2,"as":"claude-sonnet-4-6","version":"Creative Writing v3"},"eqChat":{"v":1206.5,"as":"claude-sonnet-4-6","version":"EQ-Bench 4","ciLow":1197,"ciHigh":1214.6,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":79.9,"as":"claude-sonnet-4-6","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.3,"regions":["eu","global","us"],"n":9,"tps":40,"ttft":1107,"tpsBest":45,"via":"Claude Platform on AWS","reqs":30200},"open":false,"rankable":true,"scored":13},{"id":"anthropic/claude-sonnet-4.5","name":"Anthropic: Claude Sonnet 4.5","org":"Anthropic","slug":"claude-sonnet-4.5","matchKey":"claudesonnet45","blurb":"Claude Sonnet 4.5 is Anthropic’s most advanced Sonnet model to date, optimized for real-world agents and coding workflows.","created":1759161676,"cutoff":"2025-01-31","hf":null,"ctx":1000000,"maxOut":64000,"in":["text","image","file"],"out":["text"],"price":{"in":3,"out":15,"webSearch":0.01,"alt":{"batch":{"in":1.5,"out":7.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":100,"tps":36,"ttft":1327,"reqs":3220},{"name":"Azure","region":"global","ctx":200000,"in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":98.6,"noSchema":1,"tps":38,"ttft":3731,"reqs":3},{"name":"Amazon Bedrock","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":99.9},{"name":"Anthropic","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":99.9,"tps":36,"ttft":1053,"reqs":1385},{"name":"Google","region":"global","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":99.9,"noSchema":1},{"name":"Amazon Bedrock","region":"eu","in":3.3,"out":16.5,"cacheRead":0.33,"cacheWrite":4.125,"uptime":99.7,"tps":39,"ttft":1523,"reqs":12396},{"name":"Google","region":"us","in":3.3,"out":16.5,"cacheRead":0.33,"cacheWrite":4.125,"noSchema":1,"tps":33,"ttft":1753,"reqs":72}],"scores":{"arenaChat":{"v":1456.2015004666669,"ci":2.8,"rank":58,"votes":80980,"as":"claude-sonnet-4-5-20250929-high-32k","effort":"high"},"arenaWebdev":{"v":1392.5889183050035,"ci":7.8,"rank":79,"votes":12764,"as":"claude-sonnet-4-5-20250929-high-32k","effort":"high"},"arenaDocument":{"v":1449.8080646641465,"ci":6.1,"rank":28,"votes":31455,"as":"claude-sonnet-4-5-20250929","effort":""},"arenaSearch":{"v":1158.4558124801003,"ci":5,"rank":24,"votes":127378,"as":"claude-sonnet-4-5-search","effort":""},"frontierMath":{"v":0.23859649122807017,"as":"Claude Sonnet 4.5 (32k thinking)","effort":"default","date":"2026-06-11","stderr":0.02529183228411416},"aime":{"v":0.7777777777777778,"as":"Claude Sonnet 4.5 (59k thinking)","effort":"default","date":"2025-10-28","stderr":0.06267511942419625},"gpqa":{"v":0.8232323232323232,"as":"Claude Sonnet 4.5 (59k thinking)","effort":"default","date":"2025-10-28","stderr":0.027178752639044908},"simpleqa":{"v":0.307,"as":"Claude Sonnet 4.5 (59k thinking)","effort":"default","date":"2026-08-27","stderr":0.014593284892852741},"swebench":{"v":0.7128099173553719,"as":"Claude Sonnet 4.5 (no thinking)","effort":"none","date":"2026-02-05","stderr":0.020587234937983246},"swebenchAgent":{"v":0.748,"as":"Claude 4.5 Sonnet","agent":"Sonar Foundation Agent","date":"2025-11-03","checked":false},"bfcl":{"v":0.7323999999999999,"as":"Claude-Sonnet-4-5-20250929 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1674.4,"as":"claude-sonnet-4.5","version":"Creative Writing v3"},"eqLongform":{"v":71.4,"as":"claude-sonnet-4.5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.3,"regions":["eu","global","us"],"n":7,"tps":39,"ttft":1523,"via":"Amazon Bedrock","reqs":12396},"open":false,"prose":{"slop":9.72238118855951,"trigram":0.22891152569531875,"notXButY":0.17102987605114325,"repetition":7.4341882273189475,"mattr":0.5775882041500477,"grade":6.153756138266132,"sentence":11.422189390741522,"words":157266,"samples":150,"as":"claude-sonnet-4-5"},"rankable":true,"scored":13},{"id":"openai/gpt-5.6-luna","name":"OpenAI: GPT-5.6 Luna","org":"OpenAI","slug":"gpt-5.6-luna","matchKey":"gpt56luna","blurb":"GPT-5.6 Luna is a fast, cost-efficient model in OpenAI's GPT-5.6 series.","created":1783590864,"cutoff":"2026-02-16","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":0.2,"out":1.2,"webSearch":0.01,"alt":{"batch":{"in":0.1,"out":0.6}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.1,"out":0.6,"cacheRead":0.01,"cacheWrite":0.125,"uptime":99.7},{"name":"Azure","in":0.2,"out":1.2,"cacheRead":0.02,"cacheWrite":0.25,"uptime":93.9},{"name":"OpenAI","in":0.2,"out":1.2,"cacheRead":0.02,"cacheWrite":0.25,"uptime":100},{"name":"Azure","region":"us","in":0.22,"out":1.32,"cacheRead":0.022,"cacheWrite":0.275,"uptime":100},{"name":"Amazon Bedrock","region":"us","in":0.22,"out":1.32,"cacheRead":0.022,"cacheWrite":0.275,"uptime":100,"noSchema":1,"tps":33,"ttft":510,"reqs":14897},{"name":"Azure","region":"eu","in":0.22,"out":1.32,"cacheRead":0.022,"cacheWrite":0.275,"uptime":97.5,"tps":94,"ttft":3673,"reqs":2389},{"name":"OpenAI","in":0.4,"out":2.4,"cacheRead":0.04,"cacheWrite":0.5,"uptime":100,"tps":55,"ttft":2279,"reqs":482599}],"scores":{"arenaChat":{"v":1452.4510036463553,"ci":4.8,"rank":67,"votes":28547,"as":"gpt-5.6-luna-xhigh","effort":"xhigh"},"arenaWebdev":{"v":1518.87491873604,"ci":7.6,"rank":37,"votes":7828,"as":"gpt-5.6-luna-xhigh (codex-harness)","effort":"xhigh"},"arenaVision":{"v":1258.0895147211886,"ci":8.4,"rank":40,"votes":7793,"as":"gpt-5.6-luna-xhigh","effort":"xhigh"},"arenaDocument":{"v":1456.916653229613,"ci":8.2,"rank":23,"votes":5766,"as":"gpt-5.6-luna-xhigh","effort":"xhigh"},"arenaImageCode":{"v":1503.8018296701835,"ci":12.4,"rank":28,"votes":2485,"as":"gpt-5.6-luna-xhigh (codex-harness)","effort":"xhigh"},"frontierMath":{"v":0.8210526315789474,"as":"GPT-5.6 Luna (max)","effort":"max","date":"2026-07-09","stderr":0.02274515950332741},"aime":{"v":0.9833333333333333,"as":"GPT-5.6 Luna (max)","effort":"max","date":"2026-07-09","stderr":0.011650421590325524},"gpqa":{"v":0.9160353535353535,"as":"GPT-5.6 Luna (max)","effort":"max","date":"2026-07-09","stderr":0.01728284355494035},"simpleqa":{"v":0.41,"as":"GPT-5.6 Luna (max)","effort":"max","date":"2026-08-10","stderr":0.015560917136921657},"terminal":{"v":0.1727,"as":"GPT-5.6 Luna","agent":"Codex","effort":"max","ci":0.0285,"trials":330,"cost":346.67,"secs":4088.3,"date":"2026-06-26"},"eqWriting":{"v":1825.8,"as":"gpt-5.6-luna","version":"Creative Writing v3"},"eqChat":{"v":1156.3,"as":"openai/gpt-5.6-luna","version":"EQ-Bench 4","ciLow":1144.4,"ciHigh":1170.1,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":75.5,"as":"gpt-5.6-luna","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.01,"regions":["eu","us"],"n":7,"tps":55,"ttft":2279,"tpsBest":94,"via":"OpenAI","reqs":482599},"open":false,"rankable":true,"scored":13},{"id":"openai/gpt-5.6-terra","name":"OpenAI: GPT-5.6 Terra","org":"OpenAI","slug":"gpt-5.6-terra","matchKey":"gpt56terra","blurb":"GPT-5.6 Terra is a balanced model in OpenAI's GPT-5.6 series, positioned between the flagship Sol tier and the cost-efficient Luna tier.","created":1783590857,"cutoff":"2026-02-16","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":2,"out":12,"webSearch":0.01,"alt":{"batch":{"in":1,"out":6}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":1,"out":6,"cacheRead":0.1,"cacheWrite":1.25,"uptime":100},{"name":"Azure","in":2,"out":12,"cacheRead":0.2,"cacheWrite":2.5,"uptime":98.6},{"name":"OpenAI","in":2,"out":12,"cacheRead":0.2,"cacheWrite":2.5,"uptime":99.8},{"name":"Azure","region":"us","in":2.2,"out":13.2,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100},{"name":"Amazon Bedrock","region":"us","in":2.2,"out":13.2,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100,"noSchema":1,"tps":87,"ttft":1667,"reqs":2234},{"name":"Azure","region":"eu","in":2.2,"out":13.2,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100,"tps":39,"ttft":5158,"reqs":50},{"name":"OpenAI","in":4,"out":24,"cacheRead":0.4,"cacheWrite":5,"uptime":99.9,"tps":38,"ttft":1995,"reqs":43053}],"scores":{"arenaChat":{"v":1466.0372736853974,"ci":4.8,"rank":45,"votes":28119,"as":"gpt-5.6-terra-xhigh","effort":"xhigh"},"arenaWebdev":{"v":1521.2493052131276,"ci":7.8,"rank":35,"votes":7694,"as":"gpt-5.6-terra-xhigh (codex-harness)","effort":"xhigh"},"arenaVision":{"v":1266.7396301095976,"ci":8.4,"rank":32,"votes":7726,"as":"gpt-5.6-terra-xhigh","effort":"xhigh"},"arenaDocument":{"v":1471.5811447461108,"ci":8.3,"rank":13,"votes":5787,"as":"gpt-5.6-terra-xhigh","effort":"xhigh"},"arenaImageCode":{"v":1534.5475177563917,"ci":12.6,"rank":21,"votes":2419,"as":"gpt-5.6-terra-xhigh (codex-harness)","effort":"xhigh"},"frontierMath":{"v":0.8596491228070176,"as":"GPT-5.6 Terra (max)","effort":"max","date":"2026-07-09","stderr":0.020611471958164294},"aime":{"v":0.9972222222222222,"as":"GPT-5.6 Terra (max)","effort":"max","date":"2026-07-09","stderr":0.002777777777777778},"gpqa":{"v":0.9330808080808081,"as":"GPT-5.6 Terra (max)","effort":"max","date":"2026-07-09","stderr":0.015363672147634115},"simpleqa":{"v":0.432,"as":"GPT-5.6 Terra (max)","effort":"max","date":"2026-08-10","stderr":0.015672320237336324},"terminal":{"v":0.2152,"as":"GPT-5.6 Terra","agent":"Codex","effort":"max","ci":0.0325,"trials":330,"cost":1733.52,"secs":2514.6,"date":"2026-06-26"},"eqWriting":{"v":1850.3,"as":"gpt-5.6-terra","version":"Creative Writing v3"},"eqChat":{"v":1234,"as":"openai/gpt-5.6-terra","version":"EQ-Bench 4","ciLow":1225.3,"ciHigh":1244.5,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":78,"as":"gpt-5.6-terra","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.1,"regions":["eu","us"],"n":7,"tps":38,"ttft":1995,"tpsBest":87,"via":"OpenAI","reqs":43053},"open":false,"rankable":true,"scored":13},{"id":"anthropic/claude-opus-4.5","name":"Anthropic: Claude Opus 4.5","org":"Anthropic","slug":"claude-opus-4.5","matchKey":"claudeopus45","blurb":"Claude Opus 4.5 is Anthropic’s frontier reasoning model optimized for complex software engineering, agentic workflows, and long-horizon computer use.","created":1764010580,"cutoff":null,"hf":null,"ctx":200000,"maxOut":64000,"in":["file","image","text"],"out":["text"],"price":{"in":5,"out":25,"webSearch":0.01,"alt":{"batch":{"in":2.5,"out":12.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"Claude Platform on AWS","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.8,"tps":37,"ttft":1607,"reqs":30},{"name":"Azure","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"noSchema":1,"tps":23,"ttft":1369,"reqs":1},{"name":"Google","region":"global","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.8,"noSchema":1,"tps":43,"ttft":1244,"reqs":52},{"name":"Amazon Bedrock","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.9},{"name":"Anthropic","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100,"tps":35,"ttft":1149,"reqs":106},{"name":"Amazon Bedrock","region":"eu","in":5.5,"out":27.5,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100,"tps":37,"ttft":2025,"reqs":13}],"scores":{"arenaChat":{"v":1472.7934597386325,"ci":3.9,"rank":36,"votes":36239,"as":"claude-opus-4-5-20251101-high-32k","effort":"high"},"arenaWebdev":{"v":1494.8924996935639,"ci":8.4,"rank":45,"votes":10592,"as":"claude-opus-4-5-20251101-high-32k","effort":"high"},"arenaDocument":{"v":1462.1982192559744,"ci":10.4,"rank":21,"votes":7981,"as":"claude-opus-4-5-20251101","effort":""},"arenaSearch":{"v":1179.6445104274662,"ci":5.6,"rank":19,"votes":61573,"as":"claude-opus-4-5-search","effort":""},"frontierMath":{"v":0.34385964912280703,"as":"Claude Opus 4.5 (32k thinking)","effort":"default","date":"2026-06-11","stderr":0.028185763989072587},"aime":{"v":0.8611111111111112,"as":"Claude Opus 4.5 (32k thinking)","effort":"default","date":"2025-11-24","stderr":0.042644255740946095},"gpqa":{"v":0.860479797979798,"as":"Claude Opus 4.5 (32k thinking)","effort":"default","date":"2025-11-24","stderr":0.020686959671393914},"simpleqa":{"v":0.457,"as":"Claude Opus 4.5 (32k thinking)","effort":"default","date":"2026-08-27","stderr":0.0157606915901365},"swebench":{"v":0.7665289256198347,"as":"Claude Opus 4.5 (no thinking)","effort":"none","date":"2026-02-05","stderr":0.01924895909142592},"swebenchAgent":{"v":0.792,"as":"Claude 4.5 Opus","agent":"Sonar Foundation Agent","date":"2025-12-05","checked":false},"bfcl":{"v":0.7746999999999999,"as":"Claude-Opus-4-5-20251101 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1683.2,"as":"claude-opus-4-5-20251101","version":"Creative Writing v3"},"eqLongform":{"v":73.1,"as":"claude-opus-4-5-20251101","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"regions":["eu","global"],"n":6,"tps":53,"ttft":2469,"via":"Amazon Bedrock","reqs":1408},"open":false,"rankable":true,"scored":13},{"id":"google/gemini-3-flash-preview","name":"Google: Gemini 3 Flash Preview","org":"Google","slug":"gemini-3-flash-preview","matchKey":"gemini3flashpreview","blurb":"Gemini 3 Flash Preview is a high speed, high value thinking model designed for agentic workflows, multi turn chat, and coding assistance.","created":1765987078,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","file","audio","video"],"out":["text"],"price":{"in":0.5,"out":3,"imageIn":5e-7,"audioIn":1,"webSearch":0.014,"alt":{"batch":{"in":0.25,"out":1.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Google","in":0.25,"out":1.5,"cacheRead":0.025,"cacheWrite":0.041667,"uptime":5.6},{"name":"Google AI Studio","in":0.25,"out":1.5,"cacheRead":0.025,"cacheWrite":0.041667,"uptime":99.9},{"name":"Google AI Studio","in":0.5,"out":3,"cacheRead":0.05,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"global","in":0.5,"out":3,"cacheRead":0.05,"cacheWrite":0.083333,"uptime":94.4},{"name":"Google","in":0.9,"out":5.4,"cacheRead":0.09,"cacheWrite":0.15,"uptime":99.6,"tps":65,"ttft":1211,"reqs":147906,"region":"global"},{"name":"Google AI Studio","in":0.9,"out":5.4,"cacheRead":0.09,"cacheWrite":0.15,"uptime":99.8,"tps":74,"ttft":1284,"reqs":59766}],"scores":{"arenaChat":{"v":1473.688208944633,"ci":4.4,"rank":33,"votes":30225,"as":"gemini-3-flash","effort":""},"arenaWebdev":{"v":1437.8742017225575,"ci":8.8,"rank":62,"votes":10453,"as":"gemini-3-flash","effort":""},"arenaVision":{"v":1272.3548813994935,"ci":5.4,"rank":31,"votes":36679,"as":"gemini-3-flash","effort":""},"arenaDocument":{"v":1412.994825179088,"ci":9.4,"rank":40,"votes":7158,"as":"gemini-3-flash","effort":""},"arenaImageCode":{"v":1454.0821850869145,"ci":9.4,"rank":35,"votes":7127,"as":"gemini-3-flash","effort":""},"arenaSearch":{"v":1198.0900913160758,"ci":4.7,"rank":15,"votes":149334,"as":"gemini-3-flash-grounding","effort":""},"frontierMath":{"v":0.512280701754386,"as":"Gemini 3 Flash Preview","effort":"","date":"2026-06-11","stderr":0.02966059084324674},"aime":{"v":0.9555555555555556,"as":"Gemini 3 Flash Preview","effort":"","date":"2026-08-06","stderr":0.031067790907534743},"gpqa":{"v":0.8939393939393939,"as":"Gemini 3 Flash Preview","effort":"","date":"2026-08-06","stderr":0.021938047738853106},"simpleqa":{"v":0.668,"as":"Gemini 3 Flash Preview","effort":"","date":"2026-08-27","stderr":0.014899597242811567},"swebench":{"v":0.7541322314049587,"as":"Gemini 3 Flash Preview","effort":"","date":"2026-02-18","stderr":0.019593002139508684},"swebenchAgent":{"v":0.758,"as":"Gemini 3 Flash","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"eqLongform":{"v":55.7,"as":"gemini-3-flash-preview","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.025,"regions":["global"],"n":6,"tps":65,"ttft":1211,"tpsBest":74,"via":"Google","reqs":147906},"open":false,"rankable":true,"scored":13},{"id":"google/gemini-3.5-flash","name":"Google: Gemini 3.5 Flash","org":"Google","slug":"gemini-3.5-flash","matchKey":"gemini35flash","blurb":"Gemini 3.5 Flash is Google's high-efficiency multimodal model, bringing near-Pro level coding and reasoning at Flash-tier cost and speed.","created":1779193800,"cutoff":"2025-01-01","hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":1.5,"out":9,"imageIn":0.0000015,"audioIn":3,"webSearch":0.014,"alt":{"batch":{"in":0.75,"out":4.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Google","in":0.75,"out":4.5,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":39.6},{"name":"Google AI Studio","in":0.75,"out":4.5,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":99.9},{"name":"Google","region":"global","in":1.5,"out":9,"cacheRead":0.15,"cacheWrite":0.083333,"uptime":98.7},{"name":"Google AI Studio","in":1.5,"out":9,"cacheRead":0.15,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"us","in":1.65,"out":9.9,"cacheRead":0.165,"cacheWrite":0.083333,"uptime":100},{"name":"Google","in":2.7,"out":16.2,"cacheRead":0.27,"cacheWrite":0.15,"uptime":100,"tps":87,"ttft":1459,"reqs":42821,"region":"global"},{"name":"Google AI Studio","in":2.7,"out":16.2,"cacheRead":0.27,"cacheWrite":0.15,"uptime":100,"tps":79,"ttft":1093,"reqs":4826}],"scores":{"arenaChat":{"v":1477.6813379466303,"ci":4.2,"rank":25,"votes":38257,"as":"gemini-3.5-flash-high","effort":"high"},"arenaWebdev":{"v":1500.0355056854846,"ci":7.4,"rank":44,"votes":8931,"as":"gemini-3.5-flash-high","effort":"high"},"arenaVision":{"v":1283.6995709793648,"ci":7.8,"rank":17,"votes":9579,"as":"gemini-3.5-flash-high","effort":"high"},"arenaDocument":{"v":1463.2738281558122,"ci":7.9,"rank":20,"votes":6500,"as":"gemini-3.5-flash-medium","effort":"medium"},"arenaImageCode":{"v":1499.9556413279406,"ci":12.2,"rank":30,"votes":2622,"as":"gemini-3.5-flash-medium","effort":"medium"},"frontierMath":{"v":0.6280701754385964,"as":"Gemini 3.5 Flash","effort":"","date":"2026-06-10","stderr":0.028679753751074618},"aime":{"v":0.9555555555555556,"as":"Gemini 3.5 Flash","effort":"","date":"2026-05-25","stderr":0.026696111689702044},"gpqa":{"v":0.928030303030303,"as":"Gemini 3.5 Flash","effort":"","date":"2026-05-22","stderr":0.016387002835275905},"simpleqa":{"v":0.662,"as":"Gemini 3.5 Flash","effort":"","date":"2026-08-27","stderr":0.014965960710224383},"swebench":{"v":0.7933884297520661,"as":"Gemini 3.5 Flash","effort":"","date":"2026-06-01","stderr":0.018422420250424888},"eqChat":{"v":1087.2,"as":"google/gemini-3.5-flash","version":"EQ-Bench 4","ciLow":1075.2,"ciHigh":1099.4,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":71.8,"as":"gemini-3.5-flash","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.075,"regions":["global","us"],"n":7,"tps":87,"ttft":1459,"via":"Google","reqs":42821},"open":false,"rankable":true,"scored":12},{"id":"moonshotai/kimi-k2.5","name":"MoonshotAI: Kimi K2.5","org":"MoonshotAI","slug":"kimi-k2.5","matchKey":"kimik25","blurb":"Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm.","created":1769487076,"cutoff":null,"hf":"moonshotai/Kimi-K2.5","ctx":262144,"maxOut":235929,"in":["text","image"],"out":["text"],"price":{"in":0.45,"out":2.25},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"SiliconFlow","quant":"int4","in":0.45,"out":2.25,"cacheRead":0.07,"uptime":99.9,"tps":23,"ttft":1647,"reqs":1643},{"name":"AtlasCloud","quant":"int4","in":0.49,"out":2.5,"cacheRead":0.2,"uptime":99.7,"tps":41,"ttft":792,"reqs":628},{"name":"Venice","ctx":256000,"maxOut":65536,"in":0.532,"out":3.325,"cacheRead":0.209,"uptime":98.9,"tps":96,"ttft":1369,"reqs":2021},{"name":"Novita","in":0.57,"out":2.85,"cacheRead":0.095,"uptime":99.1,"tps":31,"ttft":1679,"reqs":531},{"name":"Amazon Bedrock","region":"us","maxOut":131072,"in":0.6,"out":3,"uptime":99.6,"noSchema":1,"tps":50,"ttft":2512,"reqs":189}],"scores":{"arenaChat":{"v":1450.3504692309025,"ci":3.4,"rank":69,"votes":70513,"as":"kimi-k2.5-thinking","effort":"default"},"arenaWebdev":{"v":1436.373754984749,"ci":5.3,"rank":64,"votes":19042,"as":"kimi-k2.5-thinking","effort":"default"},"arenaVision":{"v":1249.8297199647193,"ci":5.9,"rank":46,"votes":26646,"as":"kimi-k2.5-thinking","effort":"default"},"arenaDocument":{"v":1430.0950562216417,"ci":6.7,"rank":34,"votes":21569,"as":"kimi-k2.5-thinking","effort":"default"},"arenaImageCode":{"v":1439.4529353240778,"ci":17.8,"rank":40,"votes":1531,"as":"kimi-k2.5-thinking","effort":"default"},"aime":{"v":0.922,"as":"Kimi K2.5 (Fireworks)","effort":"","date":"2026-02-02","stderr":0.039},"gpqa":{"v":0.876,"as":"Kimi K2.5 (Fireworks)","effort":"","date":"2026-02-02","stderr":0.019},"simpleqa":{"v":0.343,"as":"Kimi K2.5","effort":"","date":"2026-08-27","stderr":0.015019206922357007},"swebench":{"v":0.737603305785124,"as":"Kimi K2.5","effort":"","date":"2026-02-17","stderr":0.020017832064928274},"swebenchAgent":{"v":0.708,"as":"Kimi K2.5","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"eqWriting":{"v":1575.8,"as":"moonshotai/Kimi-K2.5","version":"Creative Writing v3"},"eqLongform":{"v":74.9,"as":"moonshotai/Kimi-K2.5","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.07,"regions":["us"],"n":5,"tps":96,"ttft":1369,"via":"Venice","reqs":2021},"open":true,"rankable":true,"scored":12},{"id":"openai/gpt-6-astra","name":"OpenAI: GPT-6 Astra","org":"OpenAI","slug":"gpt-6-astra","matchKey":"gpt6astra","blurb":"GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work.","created":1788552838,"cutoff":null,"hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":10,"out":50,"webSearch":0.01,"alt":{"batch":{"in":5,"out":25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"OpenAI","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100},{"name":"Azure","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":99.7},{"name":"OpenAI","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":100},{"name":"Azure","region":"us","in":11,"out":55,"cacheRead":1.1,"cacheWrite":13.75,"uptime":100,"tps":27,"ttft":12618,"reqs":16},{"name":"OpenAI","in":20,"out":100,"cacheRead":2,"cacheWrite":25,"uptime":100,"tps":28,"ttft":4001,"reqs":18672}],"scores":{"arenaChat":{"v":1479.7712562096513,"ci":11.7,"rank":24,"votes":2693,"as":"gpt-6-astra-max","effort":"max"},"arenaWebdev":{"v":1800.2807670457605,"ci":16.5,"rank":1,"votes":2281,"as":"gpt-6-astra-max","effort":"max"},"arenaVision":{"v":1284.0034100860937,"ci":17,"rank":16,"votes":1367,"as":"gpt-6-astra-max","effort":"max"},"arenaDocument":{"v":1467.5231106903996,"ci":14.3,"rank":16,"votes":1621,"as":"gpt-6-astra-max","effort":"max"},"arenaImageCode":{"v":1733.3447195066242,"ci":20.9,"rank":1,"votes":1085,"as":"gpt-6-astra-max","effort":"max"},"frontierMath":{"v":0.9368421052631579,"as":"GPT-6 Astra (max)","effort":"max","date":"2026-08-30","stderr":0.014434038295343787},"aime":{"v":1,"as":"GPT-6 Astra (max)","effort":"max","date":"2026-08-30","stderr":0},"gpqa":{"v":0.9577020202020202,"as":"GPT-6 Astra (max)","effort":"max","date":"2026-08-30","stderr":0.013696812097020311},"simpleqa":{"v":0.756,"as":"GPT-6 Astra (max)","effort":"max","date":"2026-08-30","stderr":0.013588548437881386},"terminal":{"v":0.5818,"as":"GPT-6 Astra","agent":"Codex","effort":"max","ci":0.0279,"trials":330,"cost":3267.18,"secs":2796.3,"date":"2026-09-03"},"eqWriting":{"v":2163.9,"as":"gpt-6-astra","version":"Creative Writing v3"},"eqLongform":{"v":82.8,"as":"gpt-6-astra","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"regions":["us"],"n":5,"tps":28,"ttft":4001,"tpsBest":34,"via":"OpenAI","reqs":18672},"open":false,"rankable":true,"scored":12},{"id":"x-ai/grok-4.5","name":"SpaceXAI: Grok 4.5","org":"SpaceXAI","slug":"grok-4.5","matchKey":"grok45","blurb":"Grok 4.5 is a model from SpaceXAI with frontier performance on coding, knowledge work, and STEM.","created":1783523154,"cutoff":null,"hf":null,"ctx":500000,"maxOut":450000,"in":["text","image","file"],"out":["text"],"price":{"in":2,"out":6,"webSearch":0.005},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"xAI","in":2,"out":6,"cacheRead":0.3,"uptime":98.9},{"name":"xAI","in":2,"out":6,"cacheRead":0.3,"uptime":97.3},{"name":"xAI","in":4,"out":12,"cacheRead":0.6,"uptime":81.4},{"name":"xAI","in":4,"out":12,"cacheRead":0.6,"uptime":87.3,"tps":48,"ttft":1230,"reqs":4358}],"scores":{"arenaChat":{"v":1468.1348973204085,"ci":4.7,"rank":42,"votes":30103,"as":"grok-4.5","effort":""},"arenaWebdev":{"v":1555.1049386593459,"ci":7.6,"rank":27,"votes":8133,"as":"grok-4.5","effort":""},"arenaVision":{"v":1279.483192366581,"ci":8.1,"rank":24,"votes":8355,"as":"grok-4.5","effort":""},"arenaDocument":{"v":1451.9590443080492,"ci":8.7,"rank":26,"votes":4965,"as":"grok-4.5","effort":""},"arenaImageCode":{"v":1563.0318171882773,"ci":12.8,"rank":15,"votes":2336,"as":"grok-4.5","effort":""},"arenaSearch":{"v":1212.646185108397,"ci":6.9,"rank":8,"votes":31505,"as":"grok-4.5","effort":""},"frontierMath":{"v":0.5719298245614035,"as":"Grok 4.5 (high)","effort":"high","date":"2026-07-09","stderr":0.029360921879029937},"aime":{"v":0.9777777777777777,"as":"Grok 4.5 (high)","effort":"high","date":"2026-07-08","stderr":0.012743145865697103},"gpqa":{"v":0.9343434343434344,"as":"Grok 4.5 (high)","effort":"high","date":"2026-07-08","stderr":0.014340395485738813},"simpleqa":{"v":0.483,"as":"Grok 4.5 (high)","effort":"high","date":"2026-08-27","stderr":0.015810153729833274},"terminal":{"v":0.1242,"as":"Grok 4.5","agent":"Grok Build","effort":"high","ci":0.0262,"trials":330,"cost":2094.11,"secs":3162.1,"date":"2026-07-16"},"eqWriting":{"v":1576,"as":"grok-4.5","version":"Creative Writing v3"}},"perf":{"uptime":98.9,"cache":0.3,"n":4,"tps":48,"ttft":1230,"via":"xAI","reqs":4358},"open":false,"rankable":true,"scored":12},{"id":"openai/gpt-5","name":"OpenAI: GPT-5","org":"OpenAI","slug":"gpt-5","matchKey":"gpt5","blurb":"GPT-5 is OpenAI’s most advanced model, offering major improvements in reasoning, code quality, and user experience.","created":1754587413,"cutoff":"2024-09-30","hf":null,"ctx":400000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":1.25,"out":10,"webSearch":0.01,"alt":{"batch":{"in":0.625,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"Azure","in":1.25,"out":10,"cacheRead":0.125,"uptime":99.9},{"name":"OpenAI","in":1.25,"out":10,"cacheRead":0.125,"uptime":100,"tps":81,"ttft":3312,"reqs":4220},{"name":"Azure","region":"eu","in":1.375,"out":11,"cacheRead":0.1375,"tps":75,"ttft":4676,"reqs":624}],"scores":{"arenaChat":{"v":1434.6549117074828,"ci":4.5,"rank":95,"votes":31428,"as":"gpt-5-high","effort":"high"},"arenaWebdev":{"v":1419.5222434984944,"ci":15.7,"rank":69,"votes":3032,"as":"gpt-5-medium","effort":"medium"},"arenaVision":{"v":1224.945593794015,"ci":7.1,"rank":64,"votes":39846,"as":"gpt-5-chat","effort":""},"arenaSearch":{"v":1132.767611287441,"ci":5.8,"rank":30,"votes":20781,"as":"gpt-5-search","effort":""},"frontierMath":{"v":0.5543859649122806,"as":"GPT-5 (high)","effort":"high","date":"2026-06-10","stderr":0.029493504108183504},"aime":{"v":0.9138888888888889,"as":"GPT-5 (high)","effort":"high","date":"2025-10-29","stderr":0.03658872186755414},"gpqa":{"v":0.8617424242424242,"as":"GPT-5 (high)","effort":"high","date":"2025-10-29","stderr":0.020787836099851995},"simpleqa":{"v":0.501,"as":"GPT-5 (high)","effort":"high","date":"2026-08-27","stderr":0.015819268290576817},"swebench":{"v":0.7355371900826446,"as":"GPT-5 (high)","effort":"high","date":"2026-02-06","stderr":0.02006832181710289},"swebenchAgent":{"v":0.7440000000000001,"as":"GPT-5","agent":"Prometheus-v1.2.1","date":"2025-10-15","checked":false},"eqWriting":{"v":1623.5,"as":"gpt-5-2025-08-07","version":"Creative Writing v3"},"eqLongform":{"v":72.8,"as":"gpt-5-2025-08-07","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.125,"regions":["eu"],"n":3,"tps":81,"ttft":3312,"via":"OpenAI","reqs":4220},"open":false,"rankable":true,"scored":12},{"id":"openai/o3","name":"OpenAI: o3","org":"OpenAI","slug":"o3","matchKey":"o3","blurb":"o3 is a well-rounded and powerful model across domains.","created":1744823457,"cutoff":"2024-06-30","hf":null,"ctx":200000,"maxOut":100000,"in":["image","text","file"],"out":["text"],"price":{"in":2,"out":8,"webSearch":0.01,"alt":{"batch":{"in":1,"out":4}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":2,"out":8,"cacheRead":0.5,"uptime":100,"tps":83,"ttft":3501,"reqs":219}],"scores":{"arenaChat":{"v":1431.551504877396,"ci":3.6,"rank":99,"votes":58579,"as":"o3-2025-04-16","effort":""},"arenaVision":{"v":1215.2365492647136,"ci":6.6,"rank":68,"votes":45554,"as":"o3-2025-04-16","effort":""},"arenaSearch":{"v":1144.2640168648277,"ci":5.5,"rank":26,"votes":20644,"as":"o3-search","effort":""},"frontierMath":{"v":0.3333333333333333,"as":"o3 (high)","effort":"high","date":"2026-08-27","stderr":0.02797271194322297},"aime":{"v":0.8444444444444444,"as":"o3 (medium)","effort":"medium","date":"2026-08-07","stderr":0.05463890236888291},"gpqa":{"v":0.8181818181818182,"as":"o3 (high)","effort":"high","date":"2025-04-16","stderr":0.021267123846387202},"simpleqa":{"v":0.494,"as":"o3 (high)","effort":"high","date":"2026-08-27","stderr":0.015818160898606833},"swebench":{"v":0.6231884057971014,"as":"o3 (medium)","effort":"medium","date":"2026-02-12","stderr":0.022072335658124498},"swebenchAgent":{"v":0.584,"as":"o3","agent":"mini-SWE-agent","date":"2025-07-26","checked":true},"bfcl":{"v":0.48560000000000003,"as":"o3-2025-04-16 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1672.5,"as":"o3","version":"Creative Writing v3"},"eqLongform":{"v":62.7,"as":"o3","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.5,"n":1,"tps":83,"ttft":3501,"via":"OpenAI","reqs":219},"open":false,"prose":{"slop":19.894018748339594,"trigram":0.22771262666514858,"notXButY":0.08584793202880854,"repetition":2.896293047384262,"mattr":0.6651397033854133,"grade":10.22110321454404,"sentence":15.298708731265872,"words":421584,"samples":295,"as":"openai/o3"},"rankable":true,"scored":12},{"id":"openai/gpt-5.1","name":"OpenAI: GPT-5.1","org":"OpenAI","slug":"gpt-5.1","matchKey":"gpt51","blurb":"GPT-5.1 is the latest frontier-grade model in the GPT-5 series, offering stronger general-purpose reasoning, improved instruction adherence, and a more natural conversational style compared to GPT-5.","created":1763060305,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["image","text","file"],"out":["text"],"price":{"in":1.25,"out":10,"webSearch":0.01,"alt":{"batch":{"in":0.625,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.625,"out":5,"cacheRead":0.0625,"uptime":100},{"name":"Azure","in":1.25,"out":10,"cacheRead":0.13,"uptime":100},{"name":"OpenAI","in":1.25,"out":10,"cacheRead":0.125,"uptime":99.9},{"name":"Azure","region":"eu","in":1.375,"out":11,"cacheRead":0.143,"tps":53,"ttft":1860,"reqs":30},{"name":"OpenAI","in":2.5,"out":20,"cacheRead":0.25,"tps":36,"ttft":1381,"reqs":8965}],"scores":{"arenaChat":{"v":1455.127511108256,"ci":3.8,"rank":64,"votes":40284,"as":"gpt-5.1-high","effort":"high"},"arenaWebdev":{"v":1392.0297717148594,"ci":11.7,"rank":80,"votes":4751,"as":"gpt-5.1-medium","effort":"medium"},"arenaVision":{"v":1250.0678105170186,"ci":8,"rank":45,"votes":9336,"as":"gpt-5.1-high","effort":"high"},"arenaDocument":{"v":1403.1671458960845,"ci":9.3,"rank":42,"votes":8244,"as":"gpt-5.1","effort":""},"arenaImageCode":{"v":1419.3453579531215,"ci":22.1,"rank":43,"votes":947,"as":"gpt-5.1-high","effort":"high"},"arenaSearch":{"v":1199.4437910533609,"ci":5.2,"rank":14,"votes":59909,"as":"gpt-5.1-search","effort":""},"aime":{"v":0.8861111111111111,"as":"GPT-5.1 (high)","effort":"high","date":"2025-11-13","stderr":0.040479727089927024},"gpqa":{"v":0.8762626262626263,"as":"GPT-5.1 (high)","effort":"high","date":"2025-11-13","stderr":0.01875368281077595},"simpleqa":{"v":0.48,"as":"GPT-5.1 (high)","effort":"high","date":"2026-08-27","stderr":0.015806639423035177},"swebench":{"v":0.6797520661157025,"as":"GPT-5.1 (high)","effort":"high","date":"2026-02-18","stderr":0.02122974956319103},"swebenchAgent":{"v":0.66,"as":"GPT 5.1","agent":"mini-SWE-agent","date":"2025-11-20","checked":false}},"perf":{"uptime":100,"cache":0.0625,"regions":["eu"],"n":5,"tps":36,"ttft":1381,"tpsBest":53,"via":"OpenAI","reqs":8965},"open":false,"rankable":true,"scored":11},{"id":"anthropic/claude-fable-5.1","name":"Anthropic: Claude Fable 5.1","org":"Anthropic","slug":"claude-fable-5.1","matchKey":"claudefable51","blurb":"Claude Fable 5.1 improves on Claude Fable 5 across the board, with the biggest gains in agentic coding, long-running agentic workflows, and knowledge work: long code refactors, front-end and visual...","created":1788285838,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":10,"out":50,"webSearch":0.01,"alt":{"batch":{"in":5,"out":25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"Azure","in":10,"out":50,"cacheRead":0.25,"cacheWrite":12.5,"uptime":99.6,"noSchema":1,"tps":45,"ttft":7516,"reqs":53},{"name":"Anthropic","in":10,"out":50,"cacheRead":0.25,"cacheWrite":12.5,"uptime":99.7,"tps":44,"ttft":6033,"reqs":10463},{"name":"Amazon Bedrock","in":10,"out":50,"cacheRead":0.25,"cacheWrite":12.5,"noSchema":1,"tps":6,"ttft":4065,"reqs":1},{"name":"Google","region":"global","in":10,"out":50,"cacheRead":0.25,"cacheWrite":12.5,"uptime":99.8,"noSchema":1,"tps":56,"ttft":3958,"reqs":1181}],"scores":{"arenaChat":{"v":1498.4730733605907,"ci":8.2,"rank":5,"votes":5783,"as":"claude-fable-5.1-max","effort":"max"},"arenaWebdev":{"v":1758.0453349337508,"ci":13.6,"rank":2,"votes":3036,"as":"claude-fable-5.1-max","effort":"max"},"arenaVision":{"v":1288.8665034700095,"ci":12.3,"rank":11,"votes":2701,"as":"claude-fable-5.1-max","effort":"max"},"arenaDocument":{"v":1512.5843185623094,"ci":15.2,"rank":2,"votes":1403,"as":"claude-fable-5.1-max","effort":"max"},"arenaImageCode":{"v":1710.3907026309043,"ci":21.3,"rank":2,"votes":964,"as":"claude-fable-5.1-max","effort":"max"},"frontierMath":{"v":0.9017543859649123,"as":"Claude Fable 5.1 (max)","effort":"max","date":"2026-09-01","stderr":0.017662067789031757},"aime":{"v":1,"as":"Claude Fable 5.1 (max)","effort":"max","date":"2026-09-01","stderr":0},"simpleqa":{"v":0.708,"as":"Claude Fable 5.1 (max)","effort":"max","date":"2026-09-01","stderr":0.014385511563477433},"terminal":{"v":0.5788,"as":"Fable 5.1","agent":"Claude Code","effort":"max","ci":0.037599999999999995,"trials":330,"cost":6243.5,"secs":3893.5,"date":"2026-09-01"},"eqWriting":{"v":2152.7,"as":"claude-fable-5-1","version":"Creative Writing v3"},"eqLongform":{"v":85.3,"as":"claude-fable-5-1","version":"Longform Writing v1.11"}},"perf":{"uptime":99.8,"cache":0.25,"regions":["global"],"n":4,"tps":44,"ttft":6033,"tpsBest":56,"via":"Anthropic","reqs":10463},"open":false,"rankable":true,"scored":11},{"id":"openai/gpt-5-mini","name":"OpenAI: GPT-5 Mini","org":"OpenAI","slug":"gpt-5-mini","matchKey":"gpt5mini","blurb":"GPT-5 Mini is a compact version of GPT-5, designed to handle lighter-weight reasoning tasks.","created":1754587407,"cutoff":"2024-05-31","hf":null,"ctx":400000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":0.25,"out":2,"webSearch":0.01,"alt":{"batch":{"in":0.125,"out":1}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.125,"out":1,"cacheRead":0.0125,"uptime":100},{"name":"Azure","maxOut":360000,"in":0.25,"out":2,"cacheRead":0.03,"uptime":100},{"name":"OpenAI","in":0.25,"out":2,"cacheRead":0.025,"uptime":100,"tps":80,"ttft":3586,"reqs":32509},{"name":"Azure","region":"eu","maxOut":360000,"in":0.275,"out":2.2,"cacheRead":0.033,"uptime":100,"tps":54,"ttft":3460,"reqs":31547}],"scores":{"arenaChat":{"v":1389.5602554314491,"ci":4.7,"rank":163,"votes":26569,"as":"gpt-5-mini-high","effort":"high"},"arenaVision":{"v":1182.7240784834655,"ci":8.2,"rank":90,"votes":28970,"as":"gpt-5-mini-high","effort":"high"},"frontierMath":{"v":0.4666666666666667,"as":"GPT-5 mini (high)","effort":"high","date":"2026-06-12","stderr":0.029603535719125714},"aime":{"v":0.8666666666666667,"as":"GPT-5 mini (high)","effort":"high","date":"2025-10-30","stderr":0.04286165199944133},"gpqa":{"v":0.75,"as":"GPT-5 mini (high)","effort":"high","date":"2025-10-30","stderr":0.022767295427113803},"simpleqa":{"v":0.216,"as":"GPT-5 mini (high)","effort":"high","date":"2026-08-10","stderr":0.013019735539307761},"swebench":{"v":0.6466942148760331,"as":"GPT-5 mini (medium)","effort":"medium","date":"2026-02-01","stderr":0.021749601139671306},"swebenchAgent":{"v":0.598,"as":"GPT 5 mini","agent":"mini-SWE-agent","date":"2025-08-07","checked":true},"bfcl":{"v":0.5546,"as":"GPT-5-mini-2025-08-07 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1310.3,"as":"gpt-5-mini-2025-08-07","version":"Creative Writing v3"},"eqLongform":{"v":56.2,"as":"gpt-5-mini-2025-08-07","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0125,"regions":["eu"],"n":4,"tps":80,"ttft":3586,"via":"OpenAI","reqs":32509},"open":false,"prose":{"slop":15.44274370057785,"trigram":0.20526676325343046,"notXButY":0.12931438507757906,"repetition":3.6240711414301376,"mattr":0.577817658697824,"grade":10.221697691158532,"sentence":17.223695988282202,"words":418967,"samples":300,"as":"openai/gpt-5-mini"},"rankable":true,"scored":11},{"id":"z-ai/glm-5.2","name":"Z.ai: GLM 5.2","org":"Z.ai","slug":"glm-5.2","matchKey":"glm52","blurb":"GLM 5.2 is a large-scale reasoning model from Z.ai.","created":1781631930,"cutoff":null,"hf":"zai-org/GLM-5.2","ctx":1048576,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.6496,"out":2.0416,"alt":{"batch":{"in":0.7,"out":2.2},"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","maxOut":163840,"in":0.5625,"out":1.8,"cacheRead":0.105,"uptime":98.8,"tps":43,"ttft":1436,"reqs":10273},{"name":"Ambient","quant":"fp8","ctx":202752,"maxOut":182476,"in":0.6,"out":2,"cacheRead":0.15,"uptime":99.5,"tps":35,"ttft":1955,"reqs":1188},{"name":"StreamLake","quant":"fp8","ctx":1024000,"maxOut":128000,"in":0.6426,"out":2.0196,"cacheRead":0.11934,"uptime":99.2,"tps":42,"ttft":1496,"reqs":14501},{"name":"Novita","quant":"fp8","in":0.6496,"out":2.0416,"cacheRead":0.12064,"uptime":99.9,"noSchema":1,"tps":69,"ttft":1916,"reqs":22322},{"name":"DigitalOcean","ctx":262144,"maxOut":235929,"in":0.7,"out":2.2,"cacheRead":0.105,"uptime":98.6,"noSchema":1,"tps":55,"ttft":855,"reqs":18313},{"name":"CoreWeave","quant":"fp4","maxOut":943718,"in":0.76,"out":2.42,"cacheRead":0.14,"uptime":100,"tps":84,"ttft":712,"reqs":17156},{"name":"AtlasCloud","quant":"fp8","in":0.938,"out":2.948,"cacheRead":0.1742,"uptime":99.6,"noSchema":1,"tps":50,"ttft":1170,"reqs":97},{"name":"Inceptron","quant":"fp4","maxOut":943718,"in":0.9563,"out":3.06,"cacheRead":0.1838,"uptime":98.9,"noSchema":1,"tps":67,"ttft":497,"reqs":1036,"region":"eu"},{"name":"Alibaba","quant":"fp8","ctx":1000000,"in":0.966,"out":3.036,"cacheRead":0.1932,"uptime":99.9},{"name":"SiliconFlow","quant":"fp8","maxOut":262144,"in":1.19,"out":3.74,"cacheRead":0.221,"uptime":100,"noSchema":1,"tps":46,"ttft":1145,"reqs":1229},{"name":"Phala","quant":"fp8","in":1.26,"out":3,"cacheRead":0.22,"uptime":99.4,"tps":53,"ttft":1778,"reqs":430},{"name":"BaseTen","quant":"fp8","maxOut":262144,"in":1.4,"out":4.4,"cacheRead":0.14,"uptime":100},{"name":"Together","ctx":512000,"maxOut":460800,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.7,"tps":64,"ttft":668,"reqs":3535},{"name":"Baidu","quant":"fp8","in":1.4,"out":4.4,"cacheRead":0.26,"uptime":100},{"name":"Wafer","maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.5,"tps":92,"ttft":575,"reqs":22831},{"name":"Fireworks","maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.14,"uptime":99.6},{"name":"BaseTen","quant":"fp8","maxOut":262144,"in":1.4,"out":4.4,"cacheRead":0.14,"uptime":100},{"name":"Venice","quant":"fp8","ctx":1000000,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.9,"tps":73,"ttft":1399,"reqs":4630},{"name":"GMICloud","quant":"fp8","maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.4,"tps":42,"ttft":2304,"reqs":48},{"name":"Parasail","quant":"fp4","ctx":262144,"maxOut":235929,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.4,"tps":78,"ttft":775,"reqs":488},{"name":"Friendli","maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.8,"tps":81,"ttft":418,"reqs":5644},{"name":"Cloudflare","ctx":262144,"maxOut":235929,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.9,"tps":33,"ttft":2122,"reqs":127},{"name":"Z.AI","quant":"fp8","in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.4,"noSchema":1,"tps":49,"ttft":2822,"reqs":3692},{"name":"BaseTen","quant":"fp8","maxOut":262144,"in":2.1,"out":6.6,"cacheRead":0.21,"uptime":100},{"name":"Fireworks","maxOut":943718,"in":2.1,"out":6.6,"cacheRead":0.21,"uptime":94.5},{"name":"BaseTen","quant":"fp8","maxOut":262144,"in":2.1,"out":6.6,"cacheRead":0.21,"uptime":100,"tps":87,"ttft":1087,"reqs":1175,"region":"us"},{"name":"Fireworks","maxOut":943718,"in":2.1,"out":6.6,"cacheRead":0.21,"uptime":97.6,"tps":148,"ttft":338,"reqs":62,"region":"us"},{"name":"Baidu","quant":"fp4","in":2.25,"out":7.88,"cacheRead":0.56,"uptime":99.9,"tps":41,"ttft":1111,"reqs":1279},{"name":"Decart","quant":"fp4","maxOut":943718,"in":2.25,"out":8,"cacheRead":0.48,"uptime":99.7,"tps":155,"ttft":866,"reqs":7892},{"name":"Alibaba","quant":"fp8","ctx":1000000,"in":2.31,"out":7.26,"cacheRead":0.462,"uptime":100,"tps":51,"ttft":1264,"reqs":837}],"scores":{"arenaChat":{"v":1472.0968104541143,"ci":4.6,"rank":38,"votes":36798,"as":"glm-5.2-max","effort":"max"},"arenaWebdev":{"v":1591.8171006512403,"ci":7.3,"rank":19,"votes":10516,"as":"glm-5.2-max","effort":"max"},"frontierMath":{"v":0.592057761732852,"as":"GLM-5.2","effort":"","date":"2026-06-19","stderr":0.029581952519606228},"aime":{"v":0.8638888888888889,"as":"GLM-5.2","effort":"","date":"2026-06-25","stderr":0.04219779330510574},"gpqa":{"v":0.9185606060606061,"as":"GLM-5.2","effort":"","date":"2026-06-24","stderr":0.016123207533687733},"simpleqa":{"v":0.342,"as":"GLM-5.2","effort":"","date":"2026-08-27","stderr":0.015008706182121802},"swebench":{"v":0.787,"as":"GLM-5.2","effort":"","date":"2026-06-25","stderr":0.018656911901656907},"eqWriting":{"v":1752.8,"as":"zai-org/GLM-5.2","version":"Creative Writing v3"},"eqChat":{"v":1221.9,"as":"z-ai/glm-5.2","version":"EQ-Bench 4","ciLow":1212.7,"ciHigh":1231.4,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":77.9,"as":"zai-org/GLM-5.2","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.105,"n":30,"tps":92,"ttft":575,"tpsBest":155,"via":"Wafer","reqs":22831,"regions":["eu","us"]},"open":true,"rankable":true,"scored":10},{"id":"moonshotai/kimi-k3","name":"MoonshotAI: Kimi K3","org":"MoonshotAI","slug":"kimi-k3","matchKey":"kimik3","blurb":"Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI.","created":1784215858,"cutoff":null,"hf":"moonshotai/Kimi-K3","ctx":1048576,"maxOut":943718,"in":["text","image","video"],"out":["text"],"price":{"in":3,"out":15},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"InferenceNet","quant":"fp4","in":1.5,"out":7.5,"cacheRead":0.15,"uptime":98.7,"tps":31,"ttft":1956,"reqs":7514},{"name":"Relace","quant":"fp4","in":1.7,"out":8.5,"cacheRead":0.17,"uptime":98.5,"noSchema":1,"tps":53,"ttft":1398,"reqs":1186},{"name":"Phala","in":2.07,"out":10.35,"cacheRead":0.207,"uptime":97.1,"tps":23,"ttft":3331,"reqs":183},{"name":"Wafer","in":2.1,"out":10.95,"cacheRead":0.21,"uptime":99.8,"tps":30,"ttft":625,"reqs":2665},{"name":"Morph","in":2.125,"out":11.9,"cacheRead":0.2465,"uptime":98.7},{"name":"Sail Research","quant":"fp4","in":2.144,"out":10.758,"cacheRead":0.244,"uptime":99.8,"tps":79,"ttft":1345,"reqs":6115},{"name":"Makora","in":2.55,"out":12.75,"cacheRead":0.256,"uptime":96.5,"tps":67,"ttft":692,"reqs":111},{"name":"DigitalOcean","in":2.55,"out":12.95,"cacheRead":0.255,"uptime":99.6,"tps":36,"ttft":1090,"reqs":171},{"name":"DeepInfra","quant":"bf16","maxOut":16384,"in":2.85,"out":14.25,"cacheRead":0.285,"uptime":96.4,"tps":44,"ttft":1792,"reqs":488},{"name":"Chutes","quant":"mxfp4","maxOut":65535,"in":3,"out":15,"cacheRead":0.3,"uptime":97.5,"noTools":1,"tps":25,"ttft":3207,"reqs":194},{"name":"Parasail","quant":"fp4","in":3,"out":15,"cacheRead":0.3,"uptime":97.8,"tps":48,"ttft":1319,"reqs":3854},{"name":"Modal","quant":"mxfp4","in":3,"out":15,"cacheRead":0.3,"uptime":99.3,"tps":60,"ttft":2170,"reqs":2676},{"name":"Together","in":3,"out":15,"cacheRead":0.3,"uptime":99.6,"tps":43,"ttft":1226,"reqs":8895},{"name":"Fireworks","in":3,"out":15,"cacheRead":0.3,"uptime":99.9},{"name":"BaseTen","quant":"fp8","maxOut":262144,"in":3,"out":15,"cacheRead":0.3,"uptime":99.8,"noSchema":1,"tps":65,"ttft":721,"reqs":1619},{"name":"Moonshot AI","quant":"mxfp4","in":3,"out":15,"cacheRead":0.3,"uptime":99.9,"tps":20,"ttft":5878,"reqs":8492},{"name":"Fireworks","region":"us","in":3.3,"out":16.5,"cacheRead":0.33,"uptime":99.6},{"name":"Alibaba","in":3.45,"out":17.25,"cacheRead":0.345,"uptime":98.2,"tps":23,"ttft":4160,"reqs":714},{"name":"Fireworks","in":4.5,"out":22.5,"cacheRead":0.45,"uptime":99.8,"noTools":1,"tps":39,"ttft":1077,"reqs":27,"region":"us"},{"name":"Morph","in":6,"out":22.5,"cacheRead":0.6,"uptime":98.3,"tps":29,"ttft":1466,"reqs":869}],"scores":{"arenaChat":{"v":1484.7659884633672,"ci":5.2,"rank":17,"votes":20987,"as":"kimi-k3-max","effort":"max"},"arenaWebdev":{"v":1674.2579751995886,"ci":11.2,"rank":5,"votes":4547,"as":"kimi-k3-max","effort":"max"},"arenaImageCode":{"v":1578.9400701319928,"ci":14,"rank":12,"votes":1999,"as":"kimi-k3-max","effort":"max"},"frontierMath":{"v":0.721830985915493,"as":"Kimi K3 (max)","effort":"max","date":"2026-07-17","stderr":0.026636607935112643},"aime":{"v":0.9722222222222222,"as":"Kimi K3 (max)","effort":"max","date":"2026-07-16","stderr":0.011158359139033182},"gpqa":{"v":0.9311868686868687,"as":"Kimi K3 (max)","effort":"max","date":"2026-07-16","stderr":0.014937225364194858},"simpleqa":{"v":0.506,"as":"Kimi K3 (max)","effort":"max","date":"2026-08-27","stderr":0.015818160898606836},"eqWriting":{"v":2070.6,"as":"kimi-k3","version":"Creative Writing v3"},"eqChat":{"v":1339.3,"as":"moonshotai/kimi-k3","version":"EQ-Bench 4","ciLow":1325,"ciHigh":1355.5,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":79.6,"as":"kimi-k3","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.15,"regions":["us"],"n":20,"tps":43,"ttft":1226,"tpsBest":79,"via":"Together","reqs":8895},"open":true,"rankable":true,"scored":10},{"id":"deepseek/deepseek-v4-pro","name":"DeepSeek: DeepSeek V4 Pro 0423","org":"DeepSeek","slug":"deepseek-v4-pro","matchKey":"deepseekv4pro","blurb":"DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window.","created":1777000679,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V4-Pro","ctx":1048576,"maxOut":384000,"in":["text"],"out":["text"],"price":{"in":0.95526,"out":1.91052},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high"]},"moderated":false,"providers":[{"name":"StreamLake","quant":"fp8","ctx":1024000,"in":0.95526,"out":1.91052,"cacheRead":0.079605,"uptime":99,"tps":36,"ttft":1901,"reqs":34618},{"name":"GMICloud","quant":"fp8","maxOut":943718,"in":0.957,"out":1.914,"cacheRead":0.07975,"uptime":98.8,"noSchema":1,"tps":28,"ttft":2493,"reqs":3007},{"name":"DigitalOcean","in":1.044,"out":2.088,"cacheRead":0.2088,"uptime":100,"noSchema":1,"tps":50,"ttft":693,"reqs":1960},{"name":"Cloudflare","maxOut":943718,"in":1.15,"out":2.55,"cacheRead":0.2,"uptime":97.2,"tps":37,"ttft":1414,"reqs":1257},{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":1.3,"out":2.6,"cacheRead":0.1,"uptime":99.4,"tps":33,"ttft":1270,"reqs":2302},{"name":"Alibaba","quant":"fp8","ctx":1000000,"maxOut":393216,"in":1.416,"out":2.832,"cacheRead":0.118,"uptime":100,"tps":48,"ttft":1610,"reqs":8754},{"name":"SiliconFlow","quant":"fp8","maxOut":393216,"in":1.50162,"out":3.135,"cacheRead":0.135,"uptime":99.4,"noSchema":1,"tps":32,"ttft":2113,"reqs":3250},{"name":"Novita","quant":"fp8","maxOut":393216,"in":1.6,"out":3.2,"cacheRead":0.135,"uptime":100,"noSchema":1,"tps":57,"ttft":1356,"reqs":13448},{"name":"Venice","ctx":1000000,"maxOut":32768,"in":1.65,"out":3.301,"cacheRead":0.33,"uptime":95.3,"tps":39,"ttft":1592,"reqs":1073},{"name":"AtlasCloud","quant":"fp4","maxOut":393216,"in":1.68,"out":3.38,"cacheRead":0.13,"uptime":99.9,"tps":32,"ttft":1248,"reqs":4576},{"name":"Baidu","quant":"fp8","maxOut":393216,"in":1.69,"out":3.38,"cacheRead":0.14,"uptime":99.4,"tps":43,"ttft":1066,"reqs":1031},{"name":"NextBit","quant":"fp8","maxOut":943718,"in":1.74,"out":3.48,"cacheRead":0.145,"uptime":98.3,"tps":42,"ttft":3734,"reqs":2628},{"name":"BaseTen","quant":"fp4","maxOut":262144,"in":1.74,"out":3.48,"cacheRead":0.145,"uptime":99.4},{"name":"BaseTen","quant":"fp4","maxOut":262144,"in":1.74,"out":3.48,"cacheRead":0.145,"uptime":99.3,"tps":112,"ttft":576,"reqs":2006,"region":"us"},{"name":"Parasail","quant":"fp8","maxOut":943718,"in":1.74,"out":3.48,"cacheRead":0.1,"uptime":99.6,"tps":65,"ttft":731,"reqs":987},{"name":"Azure","region":"us","in":1.91,"out":3.83,"cacheRead":0.16,"uptime":98.7,"noSchema":1,"tps":51,"ttft":1228,"reqs":960}],"scores":{"arenaChat":{"v":1463.3614595040572,"ci":6.8,"rank":50,"votes":9008,"as":"deepseek-v4-pro-high-20260813","effort":"high"},"arenaWebdev":{"v":1580.5321845256071,"ci":9.9,"rank":21,"votes":4258,"as":"deepseek-v4-pro-high-20260813","effort":"high"},"frontierMath":{"v":0.45263157894736844,"as":"DeepSeek v4 Pro (max)","effort":"max","date":"2026-06-17","stderr":0.029536098269922092},"aime":{"v":0.9666666666666667,"as":"DeepSeek v4 Pro (max)","effort":"max","date":"2026-06-17","stderr":0.02002208376727653},"gpqa":{"v":0.9090909090909091,"as":"DeepSeek v4 Pro (high)","effort":"high","date":"2026-08-06","stderr":0.020482086775424225},"simpleqa":{"v":0.46993987975951906,"as":"DeepSeek v4 Pro (max)","effort":"max","date":"2026-08-27","stderr":0.015806515332355317},"swebench":{"v":0.7763975155279503,"as":"DeepSeek v4 Pro (max)","effort":"max","date":"2026-06-18","stderr":0.01897828052158487},"eqWriting":{"v":1552.1,"as":"deepseek-ai/DeepSeek-V4-Pro","version":"Creative Writing v3"},"eqChat":{"v":1165.7,"as":"deepseek/deepseek-v4-pro","version":"EQ-Bench 4","ciLow":1157,"ciHigh":1175.1,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":75.6,"as":"deepseek-ai/DeepSeek-V4-Pro","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.079605,"regions":["us"],"n":16,"tps":36,"ttft":1901,"tpsBest":112,"via":"StreamLake","reqs":34618},"open":true,"rankable":true,"scored":10},{"id":"google/gemma-4-31b-it","name":"Google: Gemma 4 31B","org":"Google","slug":"gemma-4-31b-it","matchKey":"gemma431bit","blurb":"Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output.","created":1775148486,"cutoff":null,"hf":"google/gemma-4-31B-it","ctx":262144,"maxOut":16384,"in":["image","text","video"],"out":["text"],"price":{"in":0.09,"out":0.34,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","in":0.09,"out":0.34,"cacheRead":0.05,"uptime":99.1,"noTools":1},{"name":"CoreWeave","quant":"fp4","maxOut":235929,"in":0.1,"out":0.34,"cacheRead":0.1,"uptime":97.3,"tps":28,"ttft":798,"reqs":34553},{"name":"Venice","quant":"fp4","ctx":256000,"maxOut":8192,"in":0.12,"out":0.36,"cacheRead":0.09,"uptime":99.2,"tps":21,"ttft":711,"reqs":14415},{"name":"Chutes","quant":"fp4","ctx":131072,"maxOut":65536,"in":0.12,"out":0.37,"cacheRead":0.012,"uptime":94,"tps":12,"ttft":3866,"reqs":6969},{"name":"DeepInfra","quant":"fp8","in":0.13,"out":0.38,"uptime":97},{"name":"Crusoe","quant":"bf16","maxOut":262141,"in":0.14,"out":0.4,"cacheRead":0.14,"uptime":98.5,"tps":12,"ttft":754,"reqs":1470},{"name":"Friendli","maxOut":8192,"in":0.14,"out":0.4,"uptime":99.7,"tps":57,"ttft":1488,"reqs":9867},{"name":"Novita","quant":"bf16","maxOut":131072,"in":0.14,"out":0.4,"uptime":92.3,"tps":15,"ttft":3639,"reqs":1310},{"name":"Parasail","quant":"fp8","maxOut":235929,"in":0.15,"out":0.4,"cacheRead":0.06,"uptime":99.3,"tps":29,"ttft":991,"reqs":18750},{"name":"DeepInfra","quant":"fp8","ctx":131072,"maxOut":8192,"in":0.27,"out":0.76,"uptime":49.6,"noSchema":1,"tps":38,"ttft":807,"reqs":101182},{"name":"SambaNova","ctx":131072,"maxOut":117964,"in":0.38,"out":1.15,"uptime":97.4,"noSchema":1,"tps":116,"ttft":1765,"reqs":270},{"name":"Together","maxOut":235929,"in":0.39,"out":0.97},{"name":"ModelRun","quant":"fp4","maxOut":235929,"in":0.75,"out":1,"cacheRead":0.75,"uptime":100,"tps":100,"ttft":157,"reqs":21176},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.75,"out":1,"cacheRead":0.25,"uptime":86.5,"tps":21,"ttft":1378,"reqs":16907}],"scores":{"arenaChat":{"v":1451.085822407218,"ci":7.6,"rank":68,"votes":5894,"as":"gemma-4-31b","effort":""},"arenaWebdev":{"v":1364.128573967512,"ci":7,"rank":90,"votes":10139,"as":"gemma-4-31b","effort":""},"arenaVision":{"v":1261.40507576641,"ci":6.2,"rank":38,"votes":35713,"as":"gemma-4-31b","effort":""},"arenaDocument":{"v":1424.5122273663578,"ci":7.8,"rank":35,"votes":12326,"as":"gemma-4-31b","effort":""},"aime":{"v":0.7333333333333333,"as":"Gemma 4 31B IT","effort":"","date":"2026-08-06","stderr":0.06666666666666667},"gpqa":{"v":0.7575757575757576,"as":"Gemma 4 31B IT","effort":"","date":"2026-08-06","stderr":0.03053289223393202},"simpleqa":{"v":0.104,"as":"Gemma 4 31B IT","effort":"","date":"2026-08-27","stderr":0.009658016218524242},"eqWriting":{"v":1366,"as":"google/gemma-4-31B-it","version":"Creative Writing v3"},"eqChat":{"v":1120.3,"as":"google/gemma-4-31b-it","version":"EQ-Bench 4","ciLow":1109.1,"ciHigh":1133.8,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":56.5,"as":"google/gemma-4-31B-it","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.05,"n":14,"tps":38,"ttft":807,"tpsBest":116,"via":"DeepInfra","reqs":101182},"open":true,"rankable":true,"scored":10},{"id":"google/gemini-3.6-flash","name":"Google: Gemini 3.6 Flash","org":"Google","slug":"gemini-3.6-flash","matchKey":"gemini36flash","blurb":"Gemini 3.6 Flash is a high-efficiency model from Google for coding, agentic workflows, and web and app development.","created":1784646733,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.75,"out":3.75,"imageIn":7.5e-7,"audioIn":0.75,"webSearch":0.014,"alt":{"batch":{"in":0.375,"out":1.875}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Google","in":0.375,"out":1.875,"cacheRead":0.0375,"cacheWrite":0.020833,"uptime":73.3},{"name":"Google AI Studio","in":0.375,"out":1.875,"cacheRead":0.0375,"cacheWrite":0.020833,"uptime":99.9},{"name":"Google AI Studio","in":0.75,"out":3.75,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":99.8},{"name":"Google","region":"global","in":0.75,"out":3.75,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":99.7},{"name":"Google","in":1.35,"out":6.75,"cacheRead":0.135,"cacheWrite":0.075,"uptime":99.9},{"name":"Google AI Studio","in":1.35,"out":6.75,"cacheRead":0.135,"cacheWrite":0.075,"uptime":99.9,"tps":59,"ttft":1285,"reqs":1106},{"name":"Google","region":"us","in":0.825,"out":4.125,"cacheRead":0.0825,"cacheWrite":0.041667,"uptime":100,"tps":255,"ttft":19513,"reqs":17}],"scores":{"arenaChat":{"v":1479.8917260427943,"ci":4.9,"rank":23,"votes":26445,"as":"gemini-3.6-flash-high","effort":"high"},"arenaWebdev":{"v":1536.951171049402,"ci":7.7,"rank":32,"votes":8259,"as":"gemini-3.6-flash-high","effort":"high"},"arenaVision":{"v":1282.722467039345,"ci":9.7,"rank":21,"votes":5159,"as":"gemini-3.6-flash-high","effort":"high"},"arenaDocument":{"v":1456.057486273054,"ci":10.8,"rank":24,"votes":2975,"as":"gemini-3.6-flash-high","effort":"high"},"arenaImageCode":{"v":1547.3690387220258,"ci":15,"rank":16,"votes":1640,"as":"gemini-3.6-flash-high","effort":"high"},"frontierMath":{"v":0.5894736842105263,"as":"Gemini 3.6 Flash","effort":"","date":"2026-08-02","stderr":0.02919063494391404},"aime":{"v":0.9416666666666667,"as":"Gemini 3.6 Flash","effort":"","date":"2026-08-02","stderr":0.030926354359015452},"gpqa":{"v":0.9412878787878788,"as":"Gemini 3.6 Flash","effort":"","date":"2026-08-02","stderr":0.014028967942064968},"simpleqa":{"v":0.662,"as":"Gemini 3.6 Flash","effort":"","date":"2026-08-27","stderr":0.014965960710224383},"eqWriting":{"v":1599.8,"as":"gemini-3.6-flash","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.0375,"regions":["global","us"],"n":7,"tps":146,"ttft":1288,"tpsBest":255,"via":"Google","reqs":17716},"open":false,"rankable":true,"scored":10},{"id":"x-ai/grok-4.6","name":"SpaceXAI: Grok 4.6","org":"SpaceXAI","slug":"grok-4.6","matchKey":"grok46","blurb":"Grok 4.6 is a model from SpaceXAI with frontier performance on coding, knowledge work, and STEM.","created":1786548957,"cutoff":null,"hf":null,"ctx":500000,"maxOut":450000,"in":["text","image","file"],"out":["text"],"price":{"in":2,"out":6,"webSearch":0.005},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"xAI","in":2,"out":6,"cacheRead":0.5,"uptime":99.7},{"name":"xAI","in":2,"out":6,"cacheRead":0.5,"uptime":99.3},{"name":"Amazon Bedrock","region":"us","in":2.2,"out":6.6,"cacheRead":0.55,"cacheWrite":0,"uptime":100,"tps":72,"ttft":56402,"reqs":127},{"name":"xAI","in":4,"out":12,"cacheRead":1,"uptime":99.6},{"name":"xAI","in":4,"out":12,"cacheRead":1,"uptime":99.3,"tps":52,"ttft":709,"reqs":8601}],"scores":{"arenaChat":{"v":1455.5554320017718,"ci":5.8,"rank":63,"votes":15521,"as":"grok-4.6-high","effort":"high"},"arenaWebdev":{"v":1617.6400139873886,"ci":10.2,"rank":13,"votes":4282,"as":"grok-4.6-high","effort":"high"},"arenaVision":{"v":1264.5225694527956,"ci":11,"rank":35,"votes":3632,"as":"grok-4.6-high","effort":"high"},"arenaDocument":{"v":1452.3297871123095,"ci":12,"rank":25,"votes":2258,"as":"grok-4.6-high","effort":"high"},"arenaImageCode":{"v":1595.7763677671624,"ci":18.1,"rank":9,"votes":1171,"as":"grok-4.6-high","effort":"high"},"frontierMath":{"v":0.6596491228070176,"as":"Grok 4.6 (xhigh)","effort":"xhigh","date":"2026-08-14","stderr":0.028116467881852295},"aime":{"v":0.9916666666666667,"as":"Grok 4.6 (xhigh)","effort":"xhigh","date":"2026-08-14","stderr":0.004700633956814717},"gpqa":{"v":0.9400252525252525,"as":"Grok 4.6 (high)","effort":"high","date":"2026-08-12","stderr":0.014480018812911388},"simpleqa":{"v":0.493,"as":"Grok 4.6 (high)","effort":"high","date":"2026-08-27","stderr":0.01581774956184353},"terminal":{"v":0.203,"as":"Grok 4.6","agent":"Grok Build","effort":"high","ci":0.030899999999999997,"trials":330,"cost":3591.58,"secs":2351.5,"date":"2026-08-12"}},"perf":{"uptime":100,"cache":0.5,"regions":["us"],"n":5,"tps":52,"ttft":709,"tpsBest":72,"via":"xAI","reqs":8601},"open":false,"rankable":true,"scored":10},{"id":"openai/gpt-4.1","name":"OpenAI: GPT-4.1","org":"OpenAI","slug":"gpt-4.1","matchKey":"gpt41","blurb":"GPT-4.1 is a flagship large language model optimized for advanced instruction following, real-world software engineering, and long-context reasoning.","created":1744651385,"cutoff":"2024-06-30","hf":null,"ctx":1047576,"maxOut":32768,"in":["image","text","file"],"out":["text"],"price":{"in":2,"out":8,"webSearch":0.01,"alt":{"batch":{"in":1,"out":4}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","maxOut":942818,"in":2,"out":8,"cacheRead":0.5,"uptime":99.9},{"name":"OpenAI","in":2,"out":8,"cacheRead":0.5,"uptime":99.9,"tps":58,"ttft":941,"reqs":12606},{"name":"Azure","region":"eu","maxOut":942818,"in":2.2,"out":8.8,"cacheRead":0.55,"uptime":100,"tps":65,"ttft":930,"reqs":31}],"scores":{"arenaChat":{"v":1414.4730116233818,"ci":3.7,"rank":130,"votes":49941,"as":"gpt-4.1-2025-04-14","effort":""},"arenaVision":{"v":1213.8086979958202,"ci":6.7,"rank":71,"votes":41012,"as":"gpt-4.1-2025-04-14","effort":""},"frontierMath":{"v":0.05964912280701754,"as":"GPT-4.1","effort":"","date":"2026-08-27","stderr":0.014053607746337793},"aime":{"v":0.38333333333333336,"as":"GPT-4.1","effort":"","date":"2025-04-14","stderr":0.06324056203789373},"gpqa":{"v":0.6691919191919192,"as":"GPT-4.1","effort":"","date":"2025-04-14","stderr":0.02821440093855908},"simpleqa":{"v":0.311,"as":"GPT-4.1","effort":"","date":"2026-08-31","stderr":0.014645596385722692},"swebench":{"v":0.48541666666666666,"as":"GPT-4.1","effort":"","date":"2026-02-08","stderr":0.022835863746433965},"swebenchAgent":{"v":0.3958,"as":"GPT 4.1","agent":"mini-SWE-agent","date":"2025-07-26","checked":true},"bfcl":{"v":0.5396,"as":"GPT-4.1-2025-04-14 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1417.2,"as":"gpt-4.1","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.5,"regions":["eu"],"n":3,"tps":58,"ttft":941,"tpsBest":65,"via":"OpenAI","reqs":12606},"open":false,"rankable":true,"scored":10},{"id":"z-ai/glm-5.3","name":"Z.ai: GLM 5.3","org":"Z.ai","slug":"glm-5.3","matchKey":"glm53","blurb":"GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks.","created":1787086655,"cutoff":null,"hf":"zai-org/GLM-5.3","ctx":1310720,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.7,"out":2.2,"alt":{"batch":{"in":0.7,"out":2.2}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"Phala","ctx":1048576,"in":0.7,"out":2.2,"cacheRead":0.13,"uptime":99.6,"tps":37,"ttft":2221,"reqs":9433},{"name":"Novita","quant":"fp8","ctx":1048576,"in":0.7826,"out":2.4596,"cacheRead":0.14534,"uptime":88.2,"noSchema":1,"tps":33,"ttft":2814,"reqs":14465},{"name":"Reka","quant":"fp8","ctx":262144,"maxOut":235929,"in":0.7839,"out":2.6532,"cacheRead":0.15678,"uptime":96.6,"tps":40,"ttft":4130,"reqs":672},{"name":"Io Net","quant":"fp8","ctx":262124,"maxOut":32768,"in":0.806,"out":2.728,"cacheRead":0.1612,"uptime":98.5,"tps":130,"ttft":1309,"reqs":236},{"name":"InferenceNet","quant":"fp4","ctx":1000000,"in":0.9,"out":3,"cacheRead":0.15,"uptime":84.3,"tps":116,"ttft":2095,"reqs":1187},{"name":"DeepInfra","quant":"fp4","ctx":1048576,"in":0.9,"out":3,"cacheRead":0.15,"uptime":95,"tps":104,"ttft":1017,"reqs":561},{"name":"DigitalOcean","ctx":1048576,"maxOut":128000,"in":0.91,"out":2.86,"cacheRead":0.169,"uptime":64.3,"tps":49,"ttft":730,"reqs":1628},{"name":"Inceptron","quant":"fp4","ctx":1048576,"maxOut":943718,"in":0.91,"out":3.08,"cacheRead":0.182,"uptime":96.7,"noSchema":1,"tps":71,"ttft":1510,"reqs":344,"region":"eu"},{"name":"Morph","ctx":1048576,"maxOut":943718,"in":0.952,"out":2.992,"cacheRead":0.1564,"uptime":97.8,"tps":110,"ttft":712,"reqs":1254},{"name":"Sail Research","quant":"fp8","ctx":1048576,"maxOut":943718,"in":1.025,"out":3.29,"cacheRead":0.19,"uptime":94.5},{"name":"GMICloud","quant":"fp8","ctx":1048576,"maxOut":943718,"in":1.05,"out":3.3,"cacheRead":0.195,"uptime":98.9,"noSchema":1,"tps":60,"ttft":2021,"reqs":355},{"name":"Makora","quant":"fp4","ctx":980000,"maxOut":128000,"in":1.05,"out":4.2,"cacheRead":0.19,"uptime":86.4,"tps":91,"ttft":694,"reqs":1194},{"name":"SiliconFlow","quant":"fp8","ctx":1048576,"maxOut":262144,"in":1.12,"out":3.52,"cacheRead":0.208,"uptime":84.3,"noSchema":1,"tps":41,"ttft":1637,"reqs":59},{"name":"Alibaba","ctx":1000000,"in":1.19,"out":3.74,"cacheRead":0.238,"uptime":100,"noSchema":1,"tps":40,"ttft":1829,"reqs":7392},{"name":"Decart","quant":"fp4","ctx":1048576,"maxOut":943718,"in":1.19,"out":3.74,"cacheRead":0.1955,"uptime":98.2,"tps":142,"ttft":1238,"reqs":458},{"name":"Sail Research","region":"us","quant":"fp8","ctx":1048576,"maxOut":943718,"in":1.207,"out":3.871,"cacheRead":0.225,"uptime":95.3,"tps":84,"ttft":982,"reqs":578},{"name":"Friendli","ctx":1048576,"maxOut":943718,"in":1.26,"out":3.96,"cacheRead":0.234,"uptime":100,"tps":127,"ttft":500,"reqs":2506},{"name":"AkashML","quant":"fp8","ctx":1048576,"in":1.3,"out":4.4,"cacheRead":0.26,"uptime":99.9,"tps":119,"ttft":587,"reqs":299},{"name":"Baidu","quant":"fp8","ctx":1048576,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.9,"tps":42,"ttft":1552,"reqs":93},{"name":"BaseTen","quant":"fp4","ctx":1048576,"maxOut":262144,"in":1.4,"out":4.4,"cacheRead":0.14,"uptime":99.7},{"name":"Mistral","quant":"nvfp4","ctx":1048576,"in":1.4,"out":4.4,"cacheRead":0.14,"uptime":99.9,"tps":119,"ttft":1169,"reqs":6307},{"name":"Crusoe","quant":"fp4","ctx":1048576,"maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":95.3,"tps":138,"ttft":1057,"reqs":692},{"name":"Wafer","ctx":1048576,"maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":96.3,"tps":144,"ttft":540,"reqs":3020},{"name":"Venice","ctx":1000000,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":59.1,"tps":67,"ttft":3250,"reqs":256},{"name":"Together","ctx":1048575,"maxOut":943717,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":93.6,"tps":148,"ttft":332,"reqs":11160},{"name":"Parasail","quant":"fp8","ctx":1048576,"maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":98.7,"tps":127,"ttft":1021,"reqs":344},{"name":"Modal","ctx":1048576,"maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":96.9,"tps":127,"ttft":879,"reqs":5246},{"name":"BaseTen","quant":"fp4","ctx":1048576,"maxOut":262144,"in":1.4,"out":4.4,"cacheRead":0.14,"uptime":99.4},{"name":"Fireworks","ctx":1048576,"maxOut":943718,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.3,"tps":57,"ttft":1986,"reqs":417},{"name":"Cloudflare","maxOut":1179648,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.9,"tps":24,"ttft":2041,"reqs":496},{"name":"AtlasCloud","quant":"fp8","ctx":1048576,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":98.7,"tps":49,"ttft":9442,"reqs":57},{"name":"Z.AI","quant":"fp8","ctx":1048576,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.9,"noSchema":1,"tps":43,"ttft":3559,"reqs":8089},{"name":"BaseTen","quant":"fp8","ctx":1048576,"maxOut":262144,"in":2.1,"out":6.6,"cacheRead":0.21,"uptime":94.7,"noSchema":1},{"name":"BaseTen","quant":"fp8","ctx":1048576,"maxOut":262144,"in":2.1,"out":6.6,"cacheRead":0.21,"uptime":97.7,"noSchema":1,"tps":79,"ttft":1328,"reqs":180,"region":"us"}],"scores":{"arenaChat":{"v":1483.0179355440732,"ci":6.4,"rank":19,"votes":10960,"as":"glm-5.3-max","effort":"max"},"arenaWebdev":{"v":1614.4460597399911,"ci":10.8,"rank":15,"votes":3725,"as":"glm-5.3-max","effort":"max"},"frontierMath":{"v":0.6877192982456141,"as":"GLM-5.3 (max)","effort":"max","date":"2026-08-25","stderr":0.027499133473298815},"aime":{"v":0.9111111111111111,"as":"GLM-5.3 (max)","effort":"max","date":"2026-08-24","stderr":0.03454754040013868},"gpqa":{"v":0.9090909090909091,"as":"GLM-5.3 (max)","effort":"max","date":"2026-08-24","stderr":0.015851578202142853},"simpleqa":{"v":0.41,"as":"GLM-5.3 (max)","effort":"max","date":"2026-08-28","stderr":0.015560917136921659},"terminal":{"v":0.4182,"as":"GLM-5.3","agent":"Claude Code","effort":"max","ci":0.0323,"trials":330,"cost":2727.63,"secs":5831,"date":"2026-08-14"},"eqWriting":{"v":2064.1,"as":"GLM-5.3","version":"Creative Writing v3"},"eqLongform":{"v":81.8,"as":"GLM-5.3","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.13,"regions":["eu","us"],"n":34,"tps":33,"ttft":2814,"tpsBest":148,"via":"Novita","reqs":14465},"open":true,"rankable":true,"scored":9},{"id":"z-ai/glm-5.1","name":"Z.ai: GLM 5.1","org":"Z.ai","slug":"glm-5.1","matchKey":"glm51","blurb":"GLM-5.1 delivers a major leap in coding capability, with particularly significant gains in handling long-horizon tasks.","created":1775578025,"cutoff":null,"hf":"zai-org/GLM-5.1","ctx":204800,"maxOut":128000,"in":["text"],"out":["text"],"price":{"in":0.966,"out":3.036},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"StreamLake","quant":"fp8","ctx":200000,"in":0.966,"out":3.036,"cacheRead":0.1794,"uptime":99,"tps":36,"ttft":1762,"reqs":1885},{"name":"Chutes","quant":"fp8","ctx":202752,"maxOut":65535,"in":0.98,"out":3.08,"cacheRead":0.098,"uptime":81.9,"tps":26,"ttft":6685,"reqs":43},{"name":"DeepInfra","quant":"fp4","ctx":202752,"maxOut":65536,"in":1.05,"out":3.5,"cacheRead":0.205,"uptime":99.7,"tps":26,"ttft":1069,"reqs":763},{"name":"SiliconFlow","quant":"fp8","maxOut":131072,"in":1.19,"out":3.74,"cacheRead":0.6,"uptime":99.3,"noSchema":1,"tps":27,"ttft":2202,"reqs":249},{"name":"Phala","ctx":202752,"in":1.21,"out":4.2,"cacheRead":0.6,"uptime":79.6,"tps":22,"ttft":2913,"reqs":171},{"name":"AtlasCloud","quant":"fp8","ctx":202752,"maxOut":182476,"in":1.26,"out":3.96,"cacheRead":0.234,"uptime":99.5,"tps":58,"ttft":1293,"reqs":1119},{"name":"Alibaba","quant":"fp8","ctx":202745,"maxOut":131072,"in":1.33,"out":4.18,"cacheRead":0.247,"uptime":100,"tps":62,"ttft":1590,"reqs":516},{"name":"Novita","quant":"fp8","maxOut":131072,"in":1.38,"out":4.4,"cacheRead":0.26,"uptime":99.7,"noSchema":1,"tps":24,"ttft":2081,"reqs":224},{"name":"Nebius","quant":"fp8","ctx":202752,"maxOut":182476,"in":1.4,"out":4.4,"uptime":92.5,"tps":32,"ttft":947,"reqs":110},{"name":"Baidu","quant":"fp8","ctx":202752,"maxOut":131072,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.5,"tps":72,"ttft":1015,"reqs":523},{"name":"GMICloud","quant":"fp8","ctx":202752,"maxOut":182476,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":91.3,"noTools":1,"noSchema":1,"tps":66,"ttft":1684,"reqs":67},{"name":"Friendli","ctx":202752,"maxOut":182476,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":100,"tps":88,"ttft":299,"reqs":1119},{"name":"Z.AI","quant":"fp8","ctx":202752,"maxOut":131072,"in":1.4,"out":4.4,"cacheRead":0.26,"uptime":99.3,"noSchema":1,"tps":18,"ttft":8048,"reqs":2093},{"name":"Venice","quant":"fp8","ctx":200000,"maxOut":80000,"in":1.4014,"out":4.4044,"cacheRead":0.26026,"uptime":96,"tps":30,"ttft":1544,"reqs":140}],"scores":{"arenaChat":{"v":1465.5126949020796,"ci":3.9,"rank":47,"votes":48901,"as":"glm-5.1","effort":""},"arenaWebdev":{"v":1508.2915648397343,"ci":6.9,"rank":43,"votes":10220,"as":"glm-5.1","effort":""},"frontierMath":{"v":0.3684210526315789,"as":"GLM-5.1","effort":"","date":"2026-08-29","stderr":0.028623772537150215},"aime":{"v":0.9333333333333333,"as":"GLM-5.1","effort":"","date":"2026-08-10","stderr":0.03760507165451774},"gpqa":{"v":0.898989898989899,"as":"GLM-5.1","effort":"","date":"2026-08-10","stderr":0.021469735576055356},"simpleqa":{"v":0.34,"as":"GLM-5.1","effort":"","date":"2026-08-27","stderr":0.014987482264363972},"swebench":{"v":0.7417,"as":"GLM-5.1","effort":"","date":"2026-05-15","stderr":0.0199},"eqWriting":{"v":1589.2,"as":"zai-org/GLM-5.1","version":"Creative Writing v3"},"eqLongform":{"v":73.5,"as":"zai-org/GLM-5.1","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.1794,"n":14,"tps":18,"ttft":8048,"tpsBest":88,"via":"Z.AI","reqs":2093},"open":true,"rankable":true,"scored":9},{"id":"google/gemini-3.8-flash","name":"Google: Gemini 3.8 Flash","org":"Google","slug":"gemini-3.8-flash","matchKey":"gemini38flash","blurb":"Gemini 3.8 Flash is Google's most intelligent Flash model with significant gains from 3.7 Flash across software engineering, agentic tasks, and multi-step reasoning.","created":1788362056,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.75,"out":3.75,"imageIn":7.5e-7,"audioIn":0.75,"webSearch":0.014,"alt":{"batch":{"in":0.375,"out":1.875}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Google AI Studio","in":0.375,"out":1.875,"cacheRead":0.0375,"cacheWrite":0.020833,"uptime":99.9},{"name":"Google","in":0.375,"out":1.875,"cacheRead":0.0375,"cacheWrite":0.020833,"uptime":80.8},{"name":"Google AI Studio","in":0.75,"out":3.75,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":99.9},{"name":"Google","region":"global","in":0.75,"out":3.75,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":97.5},{"name":"Google AI Studio","in":1.35,"out":6.75,"cacheRead":0.135,"cacheWrite":0.075,"uptime":99.8,"tps":175,"ttft":1629,"reqs":44749},{"name":"Google","in":1.35,"out":6.75,"cacheRead":0.135,"cacheWrite":0.075,"uptime":99.9,"tps":84,"ttft":2518,"reqs":118403,"region":"global"}],"scores":{"arenaChat":{"v":1493.0082765047282,"ci":8.6,"rank":9,"votes":5076,"as":"gemini-3.8-flash-high","effort":"high"},"arenaWebdev":{"v":1567.8750817956961,"ci":12.4,"rank":23,"votes":2732,"as":"gemini-3.8-flash-high","effort":"high"},"frontierMath":{"v":0.6842105263157895,"as":"Gemini 3.8 Flash (high)","effort":"high","date":"2026-09-02","stderr":0.027582557383453006},"aime":{"v":0.9888888888888889,"as":"Gemini 3.8 Flash (high)","effort":"high","date":"2026-09-02","stderr":0.008724007409153888},"gpqa":{"v":0.9539141414141414,"as":"Gemini 3.8 Flash (high)","effort":"high","date":"2026-09-02","stderr":0.01401136318210652},"simpleqa":{"v":0.697,"as":"Gemini 3.8 Flash (high)","effort":"high","date":"2026-09-02","stderr":0.014539683710535193},"terminal":{"v":0.1909,"as":"Gemini 3.8 Flash","agent":"mini-SWE-agent","effort":"high","ci":0.0336,"trials":330,"cost":1828.77,"secs":1995.1,"date":"2026-09-02"},"eqWriting":{"v":1749.7,"as":"gemini-3.8-flash","version":"Creative Writing v3"},"eqLongform":{"v":76.8,"as":"gemini-3.8-flash","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.0375,"regions":["global"],"n":6,"tps":84,"ttft":2518,"tpsBest":175,"via":"Google","reqs":118403},"open":false,"rankable":true,"scored":9},{"id":"openai/gpt-5.4-mini","name":"OpenAI: GPT-5.4 Mini","org":"OpenAI","slug":"gpt-5.4-mini","matchKey":"gpt54mini","blurb":"GPT-5.4 mini brings the core capabilities of GPT-5.4 to a faster, more efficient model optimized for high-throughput workloads.","created":1773748178,"cutoff":"2025-08-31","hf":null,"ctx":400000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":0.75,"out":4.5,"webSearch":0.01,"alt":{"batch":{"in":0.375,"out":2.25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.375,"out":2.25,"cacheRead":0.0375,"uptime":99.5},{"name":"Azure","in":0.75,"out":4.5,"cacheRead":0.075,"uptime":100},{"name":"OpenAI","in":0.75,"out":4.5,"cacheRead":0.075,"uptime":99.5},{"name":"Azure","region":"us","in":0.825,"out":4.95,"cacheRead":0.0825,"uptime":100,"tps":41,"ttft":1008,"reqs":9720},{"name":"OpenAI","in":1.5,"out":9,"cacheRead":0.15,"uptime":99.8,"tps":83,"ttft":966,"reqs":33836}],"scores":{"arenaChat":{"v":1448.11363446197,"ci":3.8,"rank":74,"votes":59387,"as":"gpt-5.4-mini-high","effort":"high"},"arenaWebdev":{"v":1397.184359372688,"ci":6.7,"rank":78,"votes":11279,"as":"gpt-5.4-mini-high","effort":"high"},"arenaVision":{"v":1251.5303855160212,"ci":6.5,"rank":43,"votes":24041,"as":"gpt-5.4-mini-high","effort":"high"},"frontierMath":{"v":0.512280701754386,"as":"GPT-5.4 mini (xhigh)","effort":"xhigh","date":"2026-06-12","stderr":0.02966059084324674},"aime":{"v":0.8888888888888888,"as":"GPT-5.4 mini (xhigh)","effort":"xhigh","date":"2026-08-07","stderr":0.047377936967913425},"gpqa":{"v":0.8686868686868687,"as":"GPT-5.4 mini (xhigh)","effort":"xhigh","date":"2026-08-07","stderr":0.02406315641682249},"simpleqa":{"v":0.294,"as":"GPT-5.4 mini (high)","effort":"high","date":"2026-08-27","stderr":0.014414290540008265},"eqWriting":{"v":1661.8,"as":"gpt-5.4-mini","version":"Creative Writing v3"},"eqLongform":{"v":73,"as":"gpt-5.4-mini","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0375,"regions":["us"],"n":5,"tps":83,"ttft":966,"via":"OpenAI","reqs":33836},"open":false,"rankable":true,"scored":9},{"id":"openai/gpt-5-nano","name":"OpenAI: GPT-5 Nano","org":"OpenAI","slug":"gpt-5-nano","matchKey":"gpt5nano","blurb":"GPT-5-Nano is the smallest and fastest variant in the GPT-5 system, optimized for developer tools, rapid interactions, and ultra-low latency environments.","created":1754587402,"cutoff":"2024-05-31","hf":null,"ctx":400000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":0.05,"out":0.4,"webSearch":0.01,"alt":{"batch":{"in":0.025,"out":0.2}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.025,"out":0.2,"cacheRead":0.0025,"uptime":100},{"name":"Azure","maxOut":360000,"in":0.05,"out":0.4,"cacheRead":0.01,"uptime":100},{"name":"OpenAI","in":0.05,"out":0.4,"cacheRead":0.005,"uptime":100,"tps":97,"ttft":1051,"reqs":53789},{"name":"Azure","region":"eu","maxOut":360000,"in":0.055,"out":0.44,"cacheRead":0.011,"uptime":100,"tps":68,"ttft":1069,"reqs":3779}],"scores":{"arenaVision":{"v":1145.7991171073782,"ci":11.4,"rank":105,"votes":3948,"as":"gpt-5-nano-high","effort":"high"},"frontierMath":{"v":0.2,"as":"GPT-5 nano (high)","effort":"high","date":"2026-06-12","stderr":0.023735633163877067},"aime":{"v":0.8111111111111111,"as":"GPT-5 nano (high)","effort":"high","date":"2025-10-31","stderr":0.04626556218905063},"gpqa":{"v":0.6944444444444444,"as":"GPT-5 nano (high)","effort":"high","date":"2025-10-30","stderr":0.027594508998251045},"simpleqa":{"v":0.117,"as":"GPT-5 nano (high)","effort":"high","date":"2026-08-10","stderr":0.010169287802713343},"swebenchAgent":{"v":0.348,"as":"GPT 5 nano","agent":"mini-SWE-agent","date":"2025-08-07","checked":true},"bfcl":{"v":0.5145000000000001,"as":"GPT-5-nano-2025-08-07 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":702.9,"as":"gpt-5-nano-2025-08-07","version":"Creative Writing v3"},"eqLongform":{"v":28,"as":"gpt-5-nano-2025-08-07","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0025,"regions":["eu"],"n":4,"tps":97,"ttft":1051,"via":"OpenAI","reqs":53789},"open":false,"prose":{"slop":16.923060030086415,"trigram":0.28768762833393724,"notXButY":0.302556767349028,"repetition":5.399391595932562,"mattr":0.5537887375343836,"grade":12.615519786457387,"sentence":24.35788250211327,"words":455355,"samples":300,"as":"openai/gpt-5-nano"},"rankable":true,"scored":9},{"id":"thinkingmachines/inkling","name":"Thinking Machines: Inkling","org":"Thinking Machines","slug":"inkling","matchKey":"inkling","blurb":"Inkling is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 41B active parameters out of 975B total.","created":1784325956,"cutoff":null,"hf":"thinkingmachines/Inkling","ctx":1048576,"maxOut":471859,"in":["text","image","audio"],"out":["text"],"price":{"in":1,"out":4.05,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","medium","low","minimal","none"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","ctx":524288,"maxOut":262144,"in":0.95,"out":4.05,"cacheRead":0.16,"uptime":38.7,"tps":157,"ttft":448,"reqs":11},{"name":"BaseTen","quant":"fp8","maxOut":32768,"in":1,"out":4.05,"cacheRead":0.17,"uptime":99.9},{"name":"Together","ctx":524288,"in":1,"out":4.05,"cacheRead":0.17,"uptime":99.7,"tps":88,"ttft":165,"reqs":631},{"name":"BaseTen","quant":"fp8","maxOut":32768,"in":1,"out":4.05,"cacheRead":0.17,"uptime":99.9,"tps":170,"ttft":298,"reqs":166,"region":"us"}],"scores":{"arenaChat":{"v":1440.121967405652,"ci":4.9,"rank":85,"votes":25922,"as":"inkling","effort":""},"arenaWebdev":{"v":1409.989931049212,"ci":7.3,"rank":71,"votes":8970,"as":"inkling","effort":""},"frontierMath":{"v":0.3333333333333333,"as":"Inkling (xhigh)","effort":"xhigh","date":"2026-08-06","stderr":0.02797271194322297},"aime":{"v":0.8888888888888888,"as":"Inkling (xhigh)","effort":"xhigh","date":"2026-08-05","stderr":0.039171949640122654},"gpqa":{"v":0.8825757575757576,"as":"Inkling (xhigh)","effort":"xhigh","date":"2026-08-05","stderr":0.019355903133033585},"simpleqa":{"v":0.403,"as":"Inkling (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015518757419066513},"eqWriting":{"v":1607.8,"as":"thinkingmachines/Inkling","version":"Creative Writing v3"},"eqChat":{"v":1225.8,"as":"thinkingmachines/Inkling","version":"EQ-Bench 4","ciLow":1213.8,"ciHigh":1237.7,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":72.5,"as":"thinkingmachines/Inkling","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.16,"n":4,"tps":155,"ttft":337,"tpsBest":170,"via":"BaseTen","reqs":807,"regions":["us"]},"open":true,"rankable":true,"scored":9},{"id":"openai/gpt-4.1-mini","name":"OpenAI: GPT-4.1 Mini","org":"OpenAI","slug":"gpt-4.1-mini","matchKey":"gpt41mini","blurb":"GPT-4.1 Mini is a mid-sized model delivering performance competitive with GPT-4o at substantially lower latency and cost.","created":1744651381,"cutoff":"2024-06-30","hf":null,"ctx":1047576,"maxOut":32768,"in":["image","text","file"],"out":["text"],"price":{"in":0.4,"out":1.6,"webSearch":0.01,"alt":{"batch":{"in":0.2,"out":0.8}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","maxOut":942818,"in":0.4,"out":1.6,"cacheRead":0.1,"uptime":100},{"name":"OpenAI","in":0.4,"out":1.6,"cacheRead":0.1,"uptime":99.8,"tps":36,"ttft":918,"reqs":70382},{"name":"Azure","region":"eu","maxOut":942818,"in":0.44,"out":1.76,"cacheRead":0.11,"uptime":100,"tps":35,"ttft":1077,"reqs":1321}],"scores":{"arenaChat":{"v":1382.515963639083,"ci":4.3,"rank":172,"votes":38631,"as":"gpt-4.1-mini-2025-04-14","effort":""},"arenaVision":{"v":1200.8475926470985,"ci":7.5,"rank":78,"votes":40274,"as":"gpt-4.1-mini-2025-04-14","effort":""},"frontierMath":{"v":0.06666666666666667,"as":"GPT-4.1 mini","effort":"","date":"2026-08-27","stderr":0.014801767859562859},"aime":{"v":0.44722222222222224,"as":"GPT-4.1 mini","effort":"","date":"2025-04-14","stderr":0.06148017573965347},"gpqa":{"v":0.6584595959595959,"as":"GPT-4.1 mini","effort":"","date":"2025-04-14","stderr":0.027051500965295915},"simpleqa":{"v":0.127,"as":"GPT-4.1 mini","effort":"","date":"2026-08-31","stderr":0.010534798620855644},"swebenchAgent":{"v":0.2394,"as":"GPT 4.1 mini","agent":"mini-SWE-agent","date":"2025-07-20","checked":true},"bfcl":{"v":0.5045000000000001,"as":"GPT-4.1-mini-2025-04-14 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1144.7,"as":"gpt-4.1-mini","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.1,"regions":["eu"],"n":3,"tps":36,"ttft":918,"tpsBest":41,"via":"OpenAI","reqs":70382},"open":false,"rankable":true,"scored":9},{"id":"thinkingmachines/inkling-small","name":"Thinking Machines: Inkling Small","org":"Thinking Machines","slug":"inkling-small","matchKey":"inklingsmall","blurb":"Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 12B active parameters out of 276B total.","created":1785443117,"cutoff":null,"hf":"thinkingmachines/Inkling-Small","ctx":1048576,"maxOut":262144,"in":["text","image","audio"],"out":["text"],"price":{"in":0.45,"out":1.2,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","medium","low","minimal","none"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","ctx":524288,"in":0.45,"out":1.2,"cacheRead":0.1,"uptime":99.9,"tps":10,"ttft":479,"reqs":46627},{"name":"BaseTen","quant":"fp8","maxOut":32768,"in":0.5,"out":1.2,"cacheRead":0.1,"uptime":100,"tps":144,"ttft":260,"reqs":20},{"name":"Together","ctx":524288,"maxOut":471859,"in":0.5,"out":1.2,"cacheRead":0.1}],"scores":{"arenaChat":{"v":1404.608375964015,"ci":5.5,"rank":141,"votes":18844,"as":"Inkling Small","effort":""},"arenaWebdev":{"v":1406.5782695457697,"ci":9.4,"rank":73,"votes":5058,"as":"Inkling Small","effort":""},"arenaVision":{"v":1207.962969395198,"ci":9.7,"rank":73,"votes":5002,"as":"Inkling Small","effort":""},"arenaImageCode":{"v":1412.2028810182,"ci":19.8,"rank":45,"votes":970,"as":"Inkling Small","effort":""},"frontierMath":{"v":0.4631578947368421,"as":"Inkling Small (xhigh)","effort":"xhigh","date":"2026-08-14","stderr":0.02958888847874604},"aime":{"v":0.9,"as":"Inkling Small (xhigh)","effort":"xhigh","date":"2026-08-14","stderr":0.03964807305493795},"gpqa":{"v":0.88510101010101,"as":"Inkling Small (xhigh)","effort":"xhigh","date":"2026-08-14","stderr":0.019100012831823398},"simpleqa":{"v":0.191,"as":"Inkling Small (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.012436787112179482},"eqWriting":{"v":1489.6,"as":"thinkingmachines/Inkling-Small","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.1,"n":3,"tps":10,"ttft":479,"tpsBest":144,"via":"DeepInfra","reqs":46627},"open":true,"rankable":true,"scored":9},{"id":"meta/muse-spark-1.1","name":"Meta: Muse Spark 1.1","org":"Meta","slug":"muse-spark-1.1","matchKey":"musespark11","blurb":"Muse Spark 1.1 is a multimodal reasoning model from Meta, built for agentic tasks.","created":1784215741,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":943718,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":1.25,"out":4.25,"webSearch":0.0025},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"Meta","in":1.25,"out":4.25,"cacheRead":0.15,"uptime":100,"tps":221,"ttft":2598,"reqs":54}],"scores":{"arenaChat":{"v":1492.5560524214065,"ci":4.9,"rank":11,"votes":27615,"as":"muse-spark-1.1","effort":""},"arenaWebdev":{"v":1541.5474448302894,"ci":7.8,"rank":29,"votes":7549,"as":"muse-spark-1.1","effort":""},"arenaVision":{"v":1281.2462971853572,"ci":8,"rank":22,"votes":8447,"as":"muse-spark-1.1","effort":""},"arenaDocument":{"v":1464.7005226650015,"ci":8.8,"rank":18,"votes":4885,"as":"muse-spark-1.1","effort":""},"arenaImageCode":{"v":1536.6697737682216,"ci":11.7,"rank":20,"votes":2872,"as":"muse-spark-1.1","effort":""},"simpleqa":{"v":0.5778894472361809,"as":"Muse Spark 1.1","effort":"","date":"2026-08-31","stderr":0.015665429072238852},"eqWriting":{"v":1916.1,"as":"muse-spark-1.1","version":"Creative Writing v3"},"eqChat":{"v":1260.4,"as":"meta/muse-spark-1.1","version":"EQ-Bench 4","ciLow":1246.2,"ciHigh":1273.2,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":70.6,"as":"muse-spark-1.1","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.15,"n":1,"tps":221,"ttft":2598,"via":"Meta","reqs":54},"open":false,"rankable":true,"scored":9},{"id":"meta/muse-spark-1.3","name":"Meta: Muse Spark 1.3","org":"Meta","slug":"muse-spark-1.3","matchKey":"musespark13","blurb":"Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows.","created":1788378359,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":943718,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":1.25,"out":4.25,"webSearch":0.0025},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"Meta","in":1.25,"out":4.25,"cacheRead":0.15,"uptime":100,"tps":58,"ttft":3313,"reqs":38684}],"scores":{"arenaChat":{"v":1493.1148813511777,"ci":8.8,"rank":8,"votes":4723,"as":"muse-spark-1.3-max","effort":"max"},"arenaWebdev":{"v":1651.9396759222163,"ci":12.1,"rank":8,"votes":2972,"as":"muse-spark-1.3-max","effort":"max"},"arenaVision":{"v":1294.05505318626,"ci":14.9,"rank":6,"votes":1804,"as":"muse-spark-1.3-max","effort":"max"},"arenaDocument":{"v":1470.6564935613937,"ci":18.3,"rank":15,"votes":1006,"as":"muse-spark-1.3-max","effort":"max"},"arenaImageCode":{"v":1645.4797674455199,"ci":18.4,"rank":4,"votes":1152,"as":"muse-spark-1.3-max","effort":"max"},"frontierMath":{"v":0.743859649122807,"as":"Muse Spark 1.3 (xhigh)","effort":"xhigh","date":"2026-09-16","stderr":0.025901524747773286},"aime":{"v":0.9916666666666667,"as":"Muse Spark 1.3 (xhigh)","effort":"xhigh","date":"2026-09-17","stderr":0.006154574548966636},"eqWriting":{"v":1905.5,"as":"muse-spark-1.3","version":"Creative Writing v3"},"eqLongform":{"v":82.8,"as":"muse-spark-1.3","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.15,"n":1,"tps":58,"ttft":3313,"via":"Meta","reqs":38684},"open":false,"rankable":true,"scored":9},{"id":"qwen/qwen3.5-397b-a17b","name":"Qwen: Qwen3.5 397B A17B","org":"Qwen","slug":"qwen3.5-397b-a17b","matchKey":"qwen35397ba17b","blurb":"The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency.","created":1771223018,"cutoff":null,"hf":"Qwen/Qwen3.5-397B-A17B","ctx":262144,"maxOut":235929,"in":["text","image","video"],"out":["text"],"price":{"in":0.55,"out":3.5},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","maxOut":65536,"in":0.39,"out":2.34,"uptime":100,"noSchema":1,"tps":26,"ttft":1404,"reqs":1393},{"name":"DeepInfra","quant":"fp8","maxOut":81920,"in":0.45,"out":3,"cacheRead":0.22,"uptime":99.4,"tps":24,"ttft":829,"reqs":3},{"name":"Parasail","quant":"fp8","in":0.5,"out":3.6,"cacheRead":0.3,"uptime":99.7,"tps":23,"ttft":1327,"reqs":370},{"name":"DigitalOcean","ctx":131072,"maxOut":117964,"in":0.55,"out":3.5,"cacheRead":0.11,"uptime":95.4,"tps":3,"ttft":1448,"reqs":1651},{"name":"Phala","in":0.55,"out":3.5,"cacheRead":0.225,"uptime":99.6,"tps":7,"ttft":3942,"reqs":19004},{"name":"AtlasCloud","quant":"fp8","maxOut":65536,"in":0.55,"out":3.5,"cacheRead":0.55,"uptime":95.8,"tps":81,"ttft":1698,"reqs":1408},{"name":"StreamLake","ctx":256000,"maxOut":64000,"in":0.6,"out":3.6,"cacheRead":0.12,"uptime":98.7,"noTools":1,"tps":83,"ttft":1024,"reqs":1067},{"name":"GMICloud","quant":"fp8","in":0.6,"out":3.6,"uptime":89.8,"noSchema":1,"tps":15,"ttft":2455,"reqs":193},{"name":"Novita","maxOut":65536,"in":0.6,"out":3.6,"uptime":98.8,"noSchema":1,"tps":18,"ttft":2223,"reqs":566},{"name":"Venice","ctx":128000,"maxOut":32768,"in":0.75,"out":4.5,"uptime":92.5,"tps":23,"ttft":2310,"reqs":622}],"scores":{"arenaChat":{"v":1441.8293844884174,"ci":3.3,"rank":82,"votes":77007,"as":"qwen3.5-397b-a17b","effort":""},"arenaWebdev":{"v":1398.9222080160453,"ci":5.3,"rank":76,"votes":19922,"as":"qwen3.5-397b-a17b","effort":""},"arenaVision":{"v":1248.0446428346588,"ci":5.7,"rank":47,"votes":27569,"as":"qwen3.5-397b-a17b","effort":""},"frontierMath":{"v":0.312280701754386,"as":"Qwen3.5 397B-A17B","effort":"","date":"2026-08-30","stderr":0.027499133473298815},"aime":{"v":0.8888888888888888,"as":"Qwen3.5 397B-A17B","effort":"","date":"2026-08-07","stderr":0.04737793696791343},"gpqa":{"v":0.8636363636363636,"as":"Qwen3.5 397B-A17B","effort":"","date":"2026-08-07","stderr":0.02445015597318985},"eqWriting":{"v":1475.8,"as":"Qwen/Qwen3.5-397B-A17B","version":"Creative Writing v3"},"eqLongform":{"v":68.3,"as":"Qwen/Qwen3.5-397B-A17B","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":10,"tps":7,"ttft":3942,"tpsBest":83,"via":"Phala","reqs":19004},"open":true,"rankable":true,"scored":8},{"id":"google/gemini-3.5-flash-lite","name":"Google: Gemini 3.5 Flash Lite","org":"Google","slug":"gemini-3.5-flash-lite","matchKey":"gemini35flashlite","blurb":"Gemini 3.5 Flash Lite is a high-efficiency model from Google with upgraded agentic capabilities.","created":1784646726,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.3,"out":2.5,"imageIn":3e-7,"audioIn":0.3,"webSearch":0.014,"alt":{"batch":{"in":0.15,"out":1.25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Google","in":0.15,"out":1.25,"cacheRead":0.015,"cacheWrite":0.041667,"uptime":98.3},{"name":"Google AI Studio","in":0.15,"out":1.25,"cacheRead":0.015,"cacheWrite":0.041667,"uptime":100},{"name":"Google AI Studio","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":100},{"name":"Google","region":"global","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"eu","in":0.33,"out":2.75,"cacheRead":0.033,"cacheWrite":0.083333,"uptime":100},{"name":"Google","region":"us","in":0.33,"out":2.75,"cacheRead":0.033,"cacheWrite":0.083333,"uptime":100},{"name":"Google","in":0.54,"out":4.5,"cacheRead":0.054,"cacheWrite":0.15,"uptime":100,"tps":76,"ttft":2335,"reqs":7,"region":"global"},{"name":"Google AI Studio","in":0.54,"out":4.5,"cacheRead":0.054,"cacheWrite":0.15,"uptime":99.9,"tps":56,"ttft":495,"reqs":16014}],"scores":{"arenaChat":{"v":1455.8756222558854,"ci":5,"rank":59,"votes":26165,"as":"gemini-3.5-flash-lite","effort":""},"arenaWebdev":{"v":1446.6186929968917,"ci":42.5,"rank":58,"votes":228,"as":"gemini-3.5-flash-lite","effort":""},"arenaVision":{"v":1265.950576862893,"ci":9.6,"rank":34,"votes":5254,"as":"gemini-3.5-flash-lite","effort":""},"arenaImageCode":{"v":1448.3579750601916,"ci":15.9,"rank":36,"votes":1491,"as":"gemini-3.5-flash-lite","effort":""},"frontierMath":{"v":0.2596491228070175,"as":"Gemini 3.5 Flash-Lite","effort":"","date":"2026-08-02","stderr":0.02601675082238553},"aime":{"v":0.7111111111111111,"as":"Gemini 3.5 Flash-Lite","effort":"","date":"2026-08-06","stderr":0.0683294324254051},"gpqa":{"v":0.8333333333333334,"as":"Gemini 3.5 Flash-Lite","effort":"","date":"2026-08-06","stderr":0.026552207828215258},"eqWriting":{"v":1556,"as":"gemini-3.5-flash-lite","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.015,"regions":["eu","global","us"],"n":8,"tps":75,"ttft":731,"tpsBest":76,"via":"Google","reqs":71798},"open":false,"rankable":true,"scored":8},{"id":"z-ai/glm-5","name":"Z.ai: GLM 5","org":"Z.ai","slug":"glm-5","matchKey":"glm5","blurb":"GLM-5 is Z.ai’s flagship open-source foundation model engineered for complex systems design and long-horizon agent workflows.","created":1770829182,"cutoff":null,"hf":"zai-org/GLM-5","ctx":204800,"maxOut":128000,"in":["text"],"out":["text"],"price":{"in":0.6,"out":1.92},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"StreamLake","quant":"fp8","ctx":198000,"in":0.6,"out":1.92,"cacheRead":0.12,"uptime":99.8,"tps":43,"ttft":2558,"reqs":2111},{"name":"GMICloud","quant":"fp8","ctx":202752,"maxOut":182476,"in":0.6,"out":1.92,"cacheRead":0.12,"uptime":98.9,"tps":54,"ttft":1959,"reqs":871},{"name":"Baidu","quant":"fp8","ctx":202752,"maxOut":131072,"in":0.7,"out":2.24,"cacheRead":0.14,"uptime":99.5,"tps":49,"ttft":1089,"reqs":333},{"name":"SiliconFlow","quant":"fp8","maxOut":131072,"in":0.95,"out":2.55,"cacheRead":0.2,"uptime":99.7,"noSchema":1,"tps":33,"ttft":2518,"reqs":185},{"name":"Amazon Bedrock","ctx":202752,"maxOut":131072,"in":1,"out":3.2,"uptime":97.5,"noSchema":1,"tps":46,"ttft":1442,"reqs":530},{"name":"Venice","quant":"fp8","ctx":198000,"maxOut":32000,"in":1,"out":3.2,"cacheRead":0.2,"uptime":98.5,"tps":51,"ttft":1259,"reqs":70},{"name":"Novita","quant":"fp8","ctx":202800,"maxOut":131072,"in":1,"out":3.2,"cacheRead":0.2,"uptime":100,"noSchema":1,"tps":45,"ttft":1426,"reqs":278},{"name":"Z.AI","quant":"fp8","ctx":202752,"maxOut":131072,"in":1,"out":3.2,"cacheRead":0.2,"uptime":100,"noSchema":1,"tps":51,"ttft":7974,"reqs":662}],"scores":{"arenaChat":{"v":1457.6425883830184,"ci":4.3,"rank":56,"votes":27605,"as":"glm-5","effort":""},"arenaWebdev":{"v":1435.9063934463338,"ci":8.3,"rank":65,"votes":7056,"as":"glm-5","effort":""},"aime":{"v":0.8,"as":"GLM-5","effort":"","date":"2026-02-12","stderr":0.06},"gpqa":{"v":0.8781725888324873,"as":"GLM-5","effort":"","date":"2026-02-12","stderr":0.02336331210092387},"swebench":{"v":0.7207792207792207,"as":"GLM-5","effort":"","date":"2026-02-15","stderr":0.020894141602415234},"swebenchAgent":{"v":0.728,"as":"GLM 5","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"eqWriting":{"v":1597.5,"as":"zai-org/GLM-5","version":"Creative Writing v3"},"eqLongform":{"v":70.9,"as":"zai-org/GLM-5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.12,"n":8,"tps":43,"ttft":2558,"tpsBest":54,"via":"StreamLake","reqs":2111},"open":true,"rankable":true,"scored":8},{"id":"google/gemini-2.5-pro","name":"Google: Gemini 2.5 Pro","org":"Google","slug":"gemini-2.5-pro","matchKey":"gemini25pro","blurb":"Gemini 2.5 Pro is Google’s state-of-the-art AI model designed for advanced reasoning, coding, mathematics, and scientific tasks.","created":1750169544,"cutoff":"2025-01-31","hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","file","audio","video"],"out":["text"],"price":{"in":1.25,"out":10,"imageIn":0.00000125,"audioIn":1.25,"webSearch":0.014,"alt":{"batch":{"in":0.625,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Google","region":"global","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":95.9},{"name":"Google","in":2.25,"out":18,"cacheRead":0.225,"cacheWrite":0.675,"uptime":99.9},{"name":"Google AI Studio","in":0.625,"out":5,"cacheRead":0.0625,"cacheWrite":0.1875,"uptime":99.9},{"name":"Google","region":"eu","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":96.8},{"name":"Google AI Studio","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":98.3},{"name":"Google","region":"us","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":83.7,"tps":90,"ttft":3301,"reqs":11},{"name":"Google AI Studio","in":2.25,"out":18,"cacheRead":0.225,"cacheWrite":0.675,"tps":83,"ttft":4079,"reqs":552}],"scores":{"arenaChat":{"v":1445.4989968314462,"ci":2.5,"rank":77,"votes":122554,"as":"gemini-2.5-pro","effort":""},"arenaWebdev":{"v":1225.732298774946,"ci":16,"rank":121,"votes":2667,"as":"gemini-2.5-pro","effort":""},"arenaVision":{"v":1246.302652155046,"ci":4.7,"rank":48,"votes":87635,"as":"gemini-2.5-pro","effort":""},"arenaDocument":{"v":1421.2344901606084,"ci":6.3,"rank":36,"votes":25110,"as":"gemini-2.5-pro","effort":""},"arenaImageCode":{"v":1283.2861567007772,"ci":21.7,"rank":50,"votes":984,"as":"gemini-2.5-pro","effort":""},"arenaSearch":{"v":1141.6494107389728,"ci":4.6,"rank":27,"votes":83404,"as":"gemini-2.5-pro-grounding","effort":""},"gpqa":{"v":0.8383838383838383,"as":"Gemini 2.5 Pro Exp (Mar 2025)","effort":"","date":"2025-03-31","stderr":0.02622591986362927},"swebenchAgent":{"v":0.536,"as":"Gemini 2.5 Pro","agent":"mini-SWE-agent","date":"2025-07-26","checked":true}},"perf":{"uptime":99.9,"cache":0.0625,"regions":["eu","global","us"],"n":7,"tps":82,"ttft":2444,"tpsBest":101,"via":"Google","reqs":12320},"open":false,"rankable":true,"scored":8},{"id":"google/gemini-3.7-flash","name":"Google: Gemini 3.7 Flash","org":"Google","slug":"gemini-3.7-flash","matchKey":"gemini37flash","blurb":"Gemini 3.7 Flash is a multimodal model from Google for fast agentic workflows, coding, and complex multi-step reasoning.","created":1786640581,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.75,"out":3.75,"imageIn":7.5e-7,"audioIn":0.75,"webSearch":0.014,"alt":{"batch":{"in":0.375,"out":1.875}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Google","in":0.375,"out":1.875,"cacheRead":0.0375,"cacheWrite":0.020833,"uptime":85.7},{"name":"Google AI Studio","in":0.375,"out":1.875,"cacheRead":0.0375,"cacheWrite":0.020833,"uptime":99.9},{"name":"Google AI Studio","in":0.75,"out":3.75,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":99.8},{"name":"Google","region":"global","in":0.75,"out":3.75,"cacheRead":0.075,"cacheWrite":0.041667,"uptime":99.2},{"name":"Google","in":1.35,"out":6.75,"cacheRead":0.135,"cacheWrite":0.075,"uptime":99.9,"tps":93,"ttft":2377,"reqs":72914,"region":"global"},{"name":"Google AI Studio","in":1.35,"out":6.75,"cacheRead":0.135,"cacheWrite":0.075,"uptime":99.9,"tps":162,"ttft":1207,"reqs":13631}],"scores":{"arenaChat":{"v":1489.821045328777,"ci":8.2,"rank":12,"votes":5640,"as":"gemini-3.7-flash-high","effort":"high"},"arenaWebdev":{"v":1587.2627792759977,"ci":12.1,"rank":20,"votes":3007,"as":"gemini-3.7-flash-high","effort":"high"},"frontierMath":{"v":0.7157894736842105,"as":"Gemini 3.7 Flash (high)","effort":"high","date":"2026-08-14","stderr":0.026764156649364653},"aime":{"v":0.9722222222222222,"as":"Gemini 3.7 Flash (high)","effort":"high","date":"2026-08-14","stderr":0.018586784946171872},"gpqa":{"v":0.9482323232323232,"as":"Gemini 3.7 Flash (high)","effort":"high","date":"2026-08-14","stderr":0.013456525292246259},"simpleqa":{"v":0.692,"as":"Gemini 3.7 Flash (high)","effort":"high","date":"2026-08-27","stderr":0.01460648312734278},"terminal":{"v":0.1121,"as":"Gemini 3.7 Flash","agent":"mini-SWE-agent","effort":"high","ci":0.0245,"trials":330,"cost":1261.87,"secs":1688.9,"date":"2026-08-13"},"eqWriting":{"v":1722,"as":"gemini-3.7-flash","version":"Creative Writing v3"}},"perf":{"uptime":99.9,"cache":0.0375,"regions":["global"],"n":6,"tps":93,"ttft":2377,"tpsBest":162,"via":"Google","reqs":72914},"open":false,"rankable":true,"scored":8},{"id":"openai/o4-mini","name":"OpenAI: o4 Mini","org":"OpenAI","slug":"o4-mini","matchKey":"o4mini","blurb":"OpenAI o4-mini is a compact reasoning model in the o-series, optimized for fast, cost-efficient performance while retaining strong multimodal and agentic capabilities.","created":1744820942,"cutoff":"2024-06-30","hf":null,"ctx":200000,"maxOut":100000,"in":["image","text","file"],"out":["text"],"price":{"in":1.1,"out":4.4,"webSearch":0.01,"alt":{"batch":{"in":0.55,"out":2.2}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":1.1,"out":4.4,"cacheRead":0.275,"uptime":100,"tps":128,"ttft":4449,"reqs":314}],"scores":{"arenaChat":{"v":1390.814132848418,"ci":4,"rank":160,"votes":44639,"as":"o4-mini-2025-04-16","effort":""},"arenaVision":{"v":1200.3032213800766,"ci":7.2,"rank":79,"votes":41341,"as":"o4-mini-2025-04-16","effort":""},"frontierMath":{"v":0.36140350877192984,"as":"o4-mini (high)","effort":"high","date":"2026-06-11","stderr":0.028506918644975},"aime":{"v":0.8166666666666667,"as":"o4-mini (high)","effort":"high","date":"2025-04-16","stderr":0.046939139620180814},"gpqa":{"v":0.7960858585858586,"as":"o4-mini (high)","effort":"high","date":"2025-04-16","stderr":0.024034764096121958},"simpleqa":{"v":0.196,"as":"o4-mini (low)","effort":"low","date":"2026-08-27","stderr":0.012559527926707347},"swebenchAgent":{"v":0.6459999999999999,"as":"o4-mini","agent":"PatchPilot-v1.1","date":"2025-05-03","checked":false},"bfcl":{"v":0.5324,"as":"o4-mini-2025-04-16 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":100,"cache":0.275,"n":1,"tps":128,"ttft":4449,"via":"OpenAI","reqs":314},"open":false,"rankable":true,"scored":8},{"id":"qwen/qwen3.6-plus","name":"Qwen: Qwen3.6 Plus","org":"Qwen","slug":"qwen3.6-plus","matchKey":"qwen36plus","blurb":"Qwen 3.6 Plus builds on a hybrid architecture that combines efficient linear attention with sparse mixture-of-experts routing, enabling strong scalability and high-performance inference.","created":1775133557,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.325,"out":1.95},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.325,"out":1.95,"cacheWrite":0.40625,"uptime":100,"tps":34,"ttft":1033,"reqs":9594}],"scores":{"arenaChat":{"v":1443.2747768556587,"ci":4.2,"rank":79,"votes":45319,"as":"qwen3.6-plus","effort":""},"arenaWebdev":{"v":1460.5307548457781,"ci":5.7,"rank":55,"votes":17562,"as":"qwen3.6-plus","effort":""},"arenaImageCode":{"v":1468.6014466086995,"ci":10.6,"rank":34,"votes":5408,"as":"qwen3.6-plus","effort":""},"frontierMath":{"v":0.3824561403508772,"as":"Qwen 3.6 Plus (2026-04-02)","effort":"","date":"2026-08-30","stderr":0.028838027580159496},"aime":{"v":0.9333333333333333,"as":"Qwen 3.6 Plus (2026-04-02)","effort":"","date":"2026-08-07","stderr":0.03760507165451774},"gpqa":{"v":0.8838383838383839,"as":"Qwen 3.6 Plus (2026-04-02)","effort":"","date":"2026-08-07","stderr":0.022828881775249357},"simpleqa":{"v":0.44144144144144143,"as":"Qwen 3.6 Plus (2026-04-02)","effort":"","date":"2026-08-27","stderr":0.015718302238794238},"swebench":{"v":0.5785123966942148,"as":"Qwen 3.6 Plus (2026-04-02)","effort":"","date":"2026-05-14","stderr":0.02246855652298021}},"perf":{"uptime":100,"n":1,"tps":34,"ttft":1033,"via":"Alibaba","reqs":9594},"open":false,"rankable":true,"scored":8},{"id":"qwen/qwen3.7-max","name":"Qwen: Qwen3.7 Max","org":"Qwen","slug":"qwen3.7-max","matchKey":"qwen37max","blurb":"Qwen3.7-Max is the flagship model in Alibaba's Qwen3.7 series.","created":1779376861,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":1.475,"out":4.425},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":1.475,"out":4.425,"cacheRead":0.295,"cacheWrite":1.84375,"uptime":100,"tps":68,"ttft":1191,"reqs":1433}],"scores":{"arenaChat":{"v":1473.2273231258355,"ci":10,"rank":34,"votes":3705,"as":"qwen3.7-max-preview","effort":"max"},"arenaWebdev":{"v":1517.1228435798594,"ci":7.6,"rank":39,"votes":8363,"as":"qwen3.7-max-20260517","effort":"max"},"frontierMath":{"v":0.6456140350877193,"as":"Qwen3.7 Max","effort":"max","date":"2026-06-13","stderr":0.02838347520543562},"aime":{"v":0.9555555555555556,"as":"Qwen3.7 Max","effort":"max","date":"2026-08-07","stderr":0.031067790907534736},"gpqa":{"v":0.9090909090909091,"as":"Qwen3.7 Max","effort":"max","date":"2026-08-07","stderr":0.020482086775424225},"simpleqa":{"v":0.5576730190571715,"as":"Qwen3.7 Max","effort":"max","date":"2026-08-27","stderr":0.015737359397180177},"swebench":{"v":0.7727272727272727,"as":"Qwen3.7 Max","effort":"max","date":"2026-06-18","stderr":0.01906835447169652},"eqChat":{"v":1109.7,"as":"qwen/qwen3.7-max","version":"EQ-Bench 4","ciLow":1098,"ciHigh":1121.8,"boardAsOf":"2026-07-26T01:26:10+00:00"}},"perf":{"uptime":100,"cache":0.295,"n":1,"tps":68,"ttft":1191,"via":"Alibaba","reqs":1433},"open":false,"rankable":true,"scored":8},{"id":"qwen/qwen3.8-max-0902","name":"Qwen: Qwen3.8 Max (0902)","org":"Qwen","slug":"qwen3.8-max-0902","matchKey":"qwen38max0902","blurb":"Qwen3.8 Max 0902 is an updated snapshot of Qwen3.8 Max from Alibaba's Qwen team.","created":1788469704,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":2,"out":6},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Alibaba","in":2,"out":6,"cacheRead":0.25,"cacheWrite":2.5,"uptime":100,"tps":36,"ttft":2413,"reqs":9662}],"scores":{"arenaChat":{"v":1480.5610977215138,"ci":5.8,"rank":22,"votes":16670,"as":"qwen3.8-max","effort":"max"},"arenaWebdev":{"v":1680.7823383470377,"ci":14.5,"rank":4,"votes":2262,"as":"qwen3.8-max-0902","effort":"max"},"arenaVision":{"v":1301.8205065828658,"ci":8,"rank":2,"votes":8665,"as":"qwen3.8-max","effort":"max"},"arenaImageCode":{"v":1639.4396654957723,"ci":19.1,"rank":5,"votes":1094,"as":"qwen3.8-max-0902","effort":"max"},"frontierMath":{"v":0.7473684210526316,"as":"Qwen3.8 Max (xhigh)","effort":"max","date":"2026-08-04","stderr":0.0257841025556124},"aime":{"v":1,"as":"Qwen3.8 Max (0902) (xhigh)","effort":"max","date":"2026-09-02","stderr":0},"gpqa":{"v":0.9267676767676768,"as":"Qwen3.8 Max (xhigh)","effort":"max","date":"2026-08-04","stderr":0.01686498021450146},"simpleqa":{"v":0.473,"as":"Qwen3.8 Max (0902) (xhigh)","effort":"max","date":"2026-09-02","stderr":0.01579621855130273}},"perf":{"uptime":100,"cache":0.25,"n":1,"tps":36,"ttft":2413,"via":"Alibaba","reqs":9662},"open":false,"rankable":true,"scored":8},{"id":"z-ai/glm-5.3-flash","name":"Z.ai: GLM 5.3 Flash","org":"Z.ai","slug":"glm-5.3-flash","matchKey":"glm53flash","blurb":"GLM-5.3-Flash is a native multimodal model from Z.ai.","created":1787752741,"cutoff":null,"hf":"zai-org/GLM-5.3-Flash","ctx":1310720,"maxOut":943718,"in":["text","image","video"],"out":["text"],"price":{"in":0.15,"out":0.5,"alt":{"batch":{"in":0.075,"out":0.25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"OpenInference","quant":"fp4","ctx":1048576,"maxOut":102400,"in":0.075,"out":0.25,"cacheRead":0.02,"uptime":98.3,"tps":3,"ttft":5880,"reqs":33500},{"name":"GMICloud","quant":"fp8","ctx":1048576,"in":0.075,"out":0.25,"cacheRead":0.015,"uptime":98.8,"noSchema":1,"tps":20,"ttft":6890,"reqs":23018},{"name":"DeepInfra","quant":"fp4","ctx":1048576,"maxOut":131072,"in":0.075,"out":0.25,"cacheRead":0.015,"uptime":99,"tps":17,"ttft":1945,"reqs":7639},{"name":"Morph","ctx":1048576,"in":0.08,"out":0.28,"cacheRead":0.016,"uptime":99.9,"tps":17,"ttft":1714,"reqs":7196},{"name":"InferenceNet","quant":"fp4","ctx":1048576,"maxOut":128000,"in":0.09,"out":0.28,"cacheRead":0.02,"uptime":96.4,"tps":11,"ttft":2891,"reqs":22964},{"name":"Wafer","ctx":1048576,"in":0.1,"out":0.35,"cacheRead":0.02,"uptime":99.9,"tps":11,"ttft":3588,"reqs":53113},{"name":"Relace","ctx":1048576,"maxOut":131072,"in":0.11,"out":0.36,"cacheRead":0.02,"uptime":99.8,"noSchema":1,"tps":34,"ttft":1566,"reqs":129827},{"name":"Phala","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.1275,"out":0.425,"cacheRead":0.0255,"uptime":99.2,"tps":20,"ttft":3665,"reqs":35461},{"name":"Novita","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.132,"out":0.44,"cacheRead":0.0264,"uptime":99.3,"noSchema":1,"tps":26,"ttft":4615,"reqs":22800},{"name":"StreamLake","quant":"fp8","ctx":1024000,"maxOut":128000,"in":0.141,"out":0.47,"cacheRead":0.0282,"uptime":99.1,"noSchema":1,"tps":23,"ttft":2267,"reqs":30189},{"name":"Sail Research","region":"us","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.1425,"out":0.475,"cacheRead":0.0285,"uptime":98.1},{"name":"Sail Research","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.1425,"out":0.475,"cacheRead":0.0285,"uptime":99.4,"tps":24,"ttft":2023,"reqs":3687,"region":"us"},{"name":"BaseTen","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.5},{"name":"Inceptron","quant":"fp8","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":98.5,"tps":31,"ttft":1569,"reqs":1910},{"name":"Near AI","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.035,"uptime":98.1,"tps":35,"ttft":1552,"reqs":1663},{"name":"Crusoe","quant":"fp4","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":93.4,"tps":52,"ttft":1990,"reqs":1942},{"name":"CoreWeave","quant":"nvfp4","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.05,"uptime":99.7,"tps":92,"ttft":1057,"reqs":75776},{"name":"AtlasCloud","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.6,"noSchema":1,"tps":20,"ttft":4831,"reqs":9197},{"name":"Fireworks","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.6,"tps":45,"ttft":2487,"reqs":1477},{"name":"Friendli","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.4,"tps":60,"ttft":2100,"reqs":4113},{"name":"SiliconFlow","quant":"fp8","ctx":1048576,"maxOut":262144,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.5,"noSchema":1,"tps":28,"ttft":3754,"reqs":15443},{"name":"DigitalOcean","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":96.2,"tps":13,"ttft":2388,"reqs":3490},{"name":"Together","ctx":1048575,"maxOut":943717,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.3,"tps":86,"ttft":696,"reqs":61126},{"name":"Reka","quant":"fp8","ctx":262144,"maxOut":235929,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.5,"tps":11,"ttft":5179,"reqs":3750},{"name":"Parasail","quant":"fp8","ctx":1048576,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":98.9,"tps":58,"ttft":1406,"reqs":45997},{"name":"BaseTen","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.4,"tps":60,"ttft":1102,"reqs":6677,"region":"us"},{"name":"Venice","ctx":1048576,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":98.3,"tps":16,"ttft":3064,"reqs":1414},{"name":"Io Net","quant":"fp8","ctx":262144,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.4,"noSchema":1,"tps":17,"ttft":1807,"reqs":1436},{"name":"Cloudflare","maxOut":1179648,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.8,"tps":38,"ttft":1592,"reqs":5882},{"name":"Z.AI","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.15,"out":0.5,"cacheRead":0.03,"uptime":99.6,"noSchema":1,"tps":29,"ttft":4746,"reqs":123275},{"name":"NextBit","quant":"fp8","ctx":1048576,"maxOut":128000,"in":0.18,"out":0.6,"cacheRead":0.036,"uptime":97.8,"tps":31,"ttft":2859,"reqs":3871},{"name":"Modal","quant":"fp8","ctx":1048576,"in":0.45,"out":1.5,"cacheRead":0.09,"uptime":99.8,"tps":33,"ttft":290,"reqs":28737}],"scores":{"arenaChat":{"v":1475.4044535618318,"ci":6.6,"rank":29,"votes":10038,"as":"glm-5.3-flash","effort":""},"arenaWebdev":{"v":1607.092813360709,"ci":12,"rank":17,"votes":2925,"as":"glm-5.3-flash","effort":""},"arenaVision":{"v":1274.7515029907686,"ci":11.7,"rank":30,"votes":3110,"as":"glm-5.3-flash","effort":""},"arenaImageCode":{"v":1587.9397197586145,"ci":15.5,"rank":10,"votes":1600,"as":"glm-5.3-flash","effort":""},"frontierMath":{"v":0.5578947368421052,"as":"GLM-5.3-Flash (max)","effort":"max","date":"2026-08-27","stderr":0.029469977528223976},"aime":{"v":0.9388888888888889,"as":"GLM-5.3-Flash (max)","effort":"max","date":"2026-08-26","stderr":0.021622267380853318},"gpqa":{"v":0.9015151515151515,"as":"GLM-5.3-Flash (max)","effort":"max","date":"2026-08-26","stderr":0.017183231147411448}},"perf":{"uptime":99.9,"cache":0.02,"regions":["us"],"n":32,"tps":34,"ttft":1566,"tpsBest":92,"via":"Relace","reqs":129827},"open":true,"rankable":true,"scored":7},{"id":"deepseek/deepseek-v3.2","name":"DeepSeek: DeepSeek V3.2","org":"DeepSeek","slug":"deepseek-v3.2","matchKey":"deepseekv32","blurb":"DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance.","created":1764594642,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V3.2","ctx":163840,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.269,"out":0.4},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"GMICloud","quant":"fp8","maxOut":147456,"in":0.2088,"out":0.3096,"cacheRead":0.0216,"uptime":99.6,"noSchema":1,"tps":34,"ttft":2121,"reqs":3791},{"name":"StreamLake","quant":"fp8","ctx":128000,"maxOut":64000,"in":0.2145,"out":0.32175,"cacheRead":0.02145,"uptime":99.7,"tps":28,"ttft":1438,"reqs":20300},{"name":"SiliconFlow","quant":"fp8","maxOut":147456,"in":0.259,"out":0.42,"cacheRead":0.135,"uptime":98.5,"tps":13,"ttft":2614,"reqs":2756},{"name":"DeepInfra","quant":"fp4","maxOut":16384,"in":0.26,"out":0.38,"cacheRead":0.13,"uptime":99.3,"tps":12,"ttft":1048,"reqs":4311},{"name":"AtlasCloud","quant":"fp8","maxOut":147456,"in":0.26,"out":0.38,"cacheRead":0.13,"uptime":99,"tps":28,"ttft":1244,"reqs":29956},{"name":"Venice","ctx":160000,"maxOut":32768,"in":0.26829,"out":0.39024,"cacheRead":0.13008,"uptime":99.5,"tps":9,"ttft":1263,"reqs":1645},{"name":"Novita","quant":"fp8","in":0.269,"out":0.4,"cacheRead":0.1345,"uptime":100,"noSchema":1,"tps":30,"ttft":1258,"reqs":41482},{"name":"Baidu","quant":"fp8","ctx":131072,"in":0.28,"out":0.42,"cacheRead":0.028,"uptime":100,"noTools":1,"tps":32,"ttft":1503,"reqs":2706},{"name":"DigitalOcean","maxOut":147456,"in":0.3,"out":0.96,"cacheRead":0.09,"uptime":99.5,"noTools":1,"noSchema":1,"tps":16,"ttft":1166,"reqs":11420},{"name":"Alibaba","quant":"fp8","ctx":131072,"in":0.3705,"out":1.1115,"cacheRead":0.0741,"cacheWrite":0.46345,"uptime":98.7,"tps":38,"ttft":945,"reqs":1794},{"name":"Friendli","maxOut":147456,"in":0.5,"out":1.5,"cacheRead":0.25,"uptime":100,"tps":47,"ttft":435,"reqs":11881},{"name":"Google","in":0.56,"out":1.68,"uptime":99.2,"tps":17,"ttft":883,"reqs":496},{"name":"Phala","maxOut":64000,"in":1,"out":1,"cacheRead":0.5,"uptime":99.7,"tps":9,"ttft":2101,"reqs":117},{"name":"Mara","ctx":32768,"maxOut":7168,"in":3,"out":4.5,"uptime":89.2,"noTools":1,"tps":107,"ttft":1395,"reqs":38},{"name":"SambaNova","ctx":32768,"maxOut":7168,"in":3,"out":4.5,"uptime":98.3,"noTools":1,"noSchema":1,"tps":76,"ttft":1656,"reqs":57}],"scores":{"arenaChat":{"v":1425.1801092391715,"ci":3.6,"rank":111,"votes":46458,"as":"deepseek-v3.2","effort":""},"arenaWebdev":{"v":1360.5324941899548,"ci":9.2,"rank":92,"votes":6633,"as":"deepseek-v3.2-thinking","effort":"default"},"aime":{"v":0.878174603174603,"as":"DeepSeek-V3.2 (Thinking)","effort":"default","date":"2025-12-16","stderr":0.041306766974863386},"gpqa":{"v":0.8342352092352093,"as":"DeepSeek-V3.2 (Thinking)","effort":"default","date":"2025-12-16","stderr":0.0201537212608744},"swebenchAgent":{"v":0.7,"as":"DeepSeek V3.2","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"eqWriting":{"v":1511.2,"as":"deepseek-ai/DeepSeek-V3.2","version":"Creative Writing v3"},"eqLongform":{"v":66.8,"as":"deepseek-ai/DeepSeek-V3.2","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0216,"n":15,"tps":30,"ttft":1258,"tpsBest":107,"via":"Novita","reqs":41482},"open":true,"rankable":true,"scored":7},{"id":"google/gemma-4-26b-a4b-it","name":"Google: Gemma 4 26B A4B ","org":"Google","slug":"gemma-4-26b-a4b-it","matchKey":"gemma426ba4bit","blurb":"Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind.","created":1775227989,"cutoff":null,"hf":"google/gemma-4-26B-A4B-it","ctx":262144,"maxOut":235929,"in":["image","text","video"],"out":["text"],"price":{"in":0.09,"out":0.3,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Darkbloom","ctx":131072,"maxOut":32768,"in":0.042,"out":0.22,"uptime":99.9,"tps":6,"ttft":5078,"reqs":92908},{"name":"DekaLLM","quant":"bf16","in":0.06,"out":0.33,"uptime":99.9,"noSchema":1,"tps":89,"ttft":472,"reqs":28548},{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.07,"out":0.34,"uptime":99.6,"tps":22,"ttft":613,"reqs":57589},{"name":"NextBit","quant":"bf16","in":0.09,"out":0.3,"cacheRead":0.05,"uptime":98.4,"tps":49,"ttft":303,"reqs":41992},{"name":"Cloudflare","ctx":256000,"maxOut":230400,"in":0.1,"out":0.3,"uptime":99.9,"noSchema":1,"tps":48,"ttft":404,"reqs":23986},{"name":"Makora","in":0.1,"out":0.34,"cacheRead":0.034,"uptime":98.8,"noSchema":1,"tps":70,"ttft":514,"reqs":6354},{"name":"Venice","quant":"bf16","ctx":256000,"maxOut":8192,"in":0.13,"out":0.4,"cacheRead":0.05,"uptime":98.6,"tps":26,"ttft":922,"reqs":7790},{"name":"Parasail","quant":"bf16","in":0.13,"out":0.4,"cacheRead":0.05,"uptime":98.8,"noTools":1,"tps":30,"ttft":702,"reqs":5513},{"name":"Novita","quant":"bf16","maxOut":131072,"in":0.13,"out":0.4,"uptime":99.8,"noSchema":1,"tps":32,"ttft":918,"reqs":7443},{"name":"SiliconFlow","quant":"fp8","in":0.14,"out":0.4,"cacheRead":0.05,"uptime":98,"noTools":1,"tps":24,"ttft":1779,"reqs":3205},{"name":"Google","region":"global","in":0.15,"out":0.6,"uptime":97.8,"tps":21,"ttft":3057,"reqs":10924}],"scores":{"arenaChat":{"v":1437.9095676309266,"ci":7.6,"rank":88,"votes":5804,"as":"gemma-4-26b-a4b","effort":""},"arenaWebdev":{"v":1361.3953128951102,"ci":17.2,"rank":91,"votes":1442,"as":"gemma-4-26b-a4b","effort":""},"arenaVision":{"v":1241.0121540681162,"ci":6.7,"rank":52,"votes":22658,"as":"gemma-4-26b-a4b","effort":""},"aime":{"v":0.8222222222222222,"as":"Gemma 4 26B A4B","effort":"","date":"2026-08-28","stderr":0.05763774795025092},"gpqa":{"v":0.7323232323232324,"as":"Gemma 4 26B A4B","effort":"","date":"2026-08-06","stderr":0.031544498882702825},"eqWriting":{"v":1301.6,"as":"google/gemma-4-26B-A4B-it","version":"Creative Writing v3"},"eqLongform":{"v":50.7,"as":"google/gemma-4-26B-A4B-it","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"regions":["global"],"n":11,"tps":6,"ttft":5078,"tpsBest":89,"via":"Darkbloom","reqs":92908},"open":true,"rankable":true,"scored":7},{"id":"anthropic/claude-haiku-4.5","name":"Anthropic: Claude Haiku 4.5","org":"Anthropic","slug":"claude-haiku-4.5","matchKey":"claudehaiku45","blurb":"Claude Haiku 4.5 is Anthropic’s fastest and most efficient model, delivering near-frontier intelligence at a fraction of the cost and latency of larger Claude models.","created":1760547638,"cutoff":null,"hf":null,"ctx":200000,"maxOut":64000,"in":["text","image","file"],"out":["text"],"price":{"in":1,"out":5,"webSearch":0.01,"alt":{"batch":{"in":0.5,"out":2.5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"Azure","region":"global","in":1,"out":5,"cacheRead":0.1,"cacheWrite":1.25,"uptime":100,"noSchema":1,"tps":63,"ttft":552,"reqs":152},{"name":"Amazon Bedrock","region":"global","in":1,"out":5,"cacheRead":0.1,"cacheWrite":1.25,"uptime":99.9},{"name":"Google","region":"global","in":1,"out":5,"cacheRead":0.1,"cacheWrite":1.25,"uptime":99.8,"noSchema":1},{"name":"Anthropic","in":1,"out":5,"cacheRead":0.1,"cacheWrite":1.25,"uptime":100,"tps":77,"ttft":623,"reqs":13282},{"name":"Amazon Bedrock","region":"us","in":1.1,"out":5.5,"cacheRead":0.11,"cacheWrite":1.375,"uptime":100},{"name":"Google","region":"us","in":1.1,"out":5.5,"cacheRead":0.11,"cacheWrite":1.375,"uptime":100,"noSchema":1},{"name":"Amazon Bedrock","region":"eu","in":1.1,"out":5.5,"cacheRead":0.11,"cacheWrite":1.375,"uptime":100,"tps":75,"ttft":675,"reqs":70},{"name":"Google","region":"eu","in":1.1,"out":5.5,"cacheRead":0.11,"cacheWrite":1.375,"uptime":100,"noSchema":1,"tps":59,"ttft":788,"reqs":22}],"scores":{"arenaChat":{"v":1414.5523305283887,"ci":2.5,"rank":129,"votes":129278,"as":"claude-haiku-4-5-20251001","effort":""},"arenaWebdev":{"v":1329.3115918229696,"ci":5,"rank":104,"votes":27785,"as":"claude-haiku-4-5-20251001","effort":""},"arenaDocument":{"v":1419.965425258912,"ci":5.9,"rank":37,"votes":34677,"as":"claude-haiku-4-5-20251001","effort":""},"swebenchAgent":{"v":0.6659999999999999,"as":"Claude 4.5 Haiku","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"bfcl":{"v":0.687,"as":"Claude-Haiku-4-5-20251001 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqChat":{"v":1063.9,"as":"claude-haiku-4-5-20251001","version":"EQ-Bench 4","ciLow":1050.1,"ciHigh":1073.9,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":65,"as":"claude-haiku-4.5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.1,"regions":["eu","global","us"],"n":8,"tps":55,"ttft":820,"tpsBest":87,"via":"Amazon Bedrock","reqs":58245},"open":false,"rankable":true,"scored":7},{"id":"z-ai/glm-4.7","name":"Z.ai: GLM 4.7","org":"Z.ai","slug":"glm-4.7","matchKey":"glm47","blurb":"GLM-4.7 is Z.ai’s latest flagship model, featuring upgrades in two key areas: enhanced programming capabilities and more stable multi-step reasoning/execution.","created":1766378014,"cutoff":null,"hf":"zai-org/GLM-4.7","ctx":204800,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.4,"out":1.75},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","ctx":202752,"in":0.4,"out":1.75,"cacheRead":0.08,"uptime":97,"tps":19,"ttft":1019,"reqs":2687},{"name":"Venice","quant":"fp4","ctx":198000,"maxOut":16384,"in":0.4004,"out":1.9292,"cacheRead":0.08008,"uptime":89.7,"tps":8,"ttft":1240,"reqs":326},{"name":"AtlasCloud","quant":"fp8","ctx":202752,"maxOut":182476,"in":0.52,"out":1.85,"cacheRead":0.12,"uptime":91.6,"tps":38,"ttft":587,"reqs":762},{"name":"Novita","quant":"fp8","in":0.54,"out":1.98,"cacheRead":0.099,"uptime":99.3,"noSchema":1,"tps":19,"ttft":4624,"reqs":1656},{"name":"Google","ctx":200000,"maxOut":128000,"in":0.6,"out":2.2,"uptime":100,"tps":66,"ttft":499,"reqs":1571},{"name":"Z.AI","quant":"fp4","ctx":202752,"in":0.6,"out":2.2,"cacheRead":0.11,"uptime":99.1,"noSchema":1,"tps":22,"ttft":9776,"reqs":2801},{"name":"Mancer 2","quant":"fp4","ctx":131072,"maxOut":117964,"in":0.7,"out":2.5,"uptime":91,"noTools":1,"noSchema":1,"tps":8,"ttft":1431,"reqs":99}],"scores":{"arenaChat":{"v":1441.7334537056345,"ci":6.1,"rank":83,"votes":11893,"as":"glm-4.7","effort":""},"arenaWebdev":{"v":1434.7039434050089,"ci":12.4,"rank":66,"votes":3811,"as":"glm-4.7","effort":""},"aime":{"v":0.8333333333333334,"as":"GLM-4.7","effort":"","date":"2026-01-29","stderr":0.04767312946227961},"gpqa":{"v":0.8333333333333334,"as":"GLM-4.7","effort":"","date":"2026-01-29","stderr":0.02374901666263655},"simpleqa":{"v":0.322,"as":"GLM-4.7","effort":"","date":"2026-08-27","stderr":0.014782913600996744},"eqWriting":{"v":1410.8,"as":"zai-org/GLM-4.7","version":"Creative Writing v3"},"eqLongform":{"v":66,"as":"zai-org/GLM-4.7","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.08,"n":7,"tps":22,"ttft":9776,"tpsBest":66,"via":"Z.AI","reqs":2801},"open":true,"rankable":true,"scored":7},{"id":"openai/gpt-5.4-nano","name":"OpenAI: GPT-5.4 Nano","org":"OpenAI","slug":"gpt-5.4-nano","matchKey":"gpt54nano","blurb":"GPT-5.4 nano is the most lightweight and cost-efficient variant of the GPT-5.4 family, optimized for speed-critical and high-volume tasks.","created":1773748187,"cutoff":"2025-08-31","hf":null,"ctx":400000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":0.2,"out":1.25,"webSearch":0.01,"alt":{"batch":{"in":0.1,"out":0.625}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.1,"out":0.625,"cacheRead":0.01,"uptime":100},{"name":"Azure","in":0.2,"out":1.25,"cacheRead":0.02,"uptime":100},{"name":"OpenAI","in":0.2,"out":1.25,"cacheRead":0.02,"uptime":100,"tps":80,"ttft":950,"reqs":60878},{"name":"Azure","region":"us","in":0.22,"out":1.375,"cacheRead":0.022,"tps":39,"ttft":1973,"reqs":5585}],"scores":{"arenaChat":{"v":1402.0797629981867,"ci":3.9,"rank":146,"votes":58424,"as":"gpt-5.4-nano-high","effort":"high"},"arenaVision":{"v":1201.257270964332,"ci":6.6,"rank":76,"votes":23906,"as":"gpt-5.4-nano-high","effort":"high"},"frontierMath":{"v":0.44912280701754387,"as":"GPT-5.4 nano (high)","effort":"high","date":"2026-06-12","stderr":0.029515543245521678},"aime":{"v":0.8778,"as":"GPT-5.4 nano (high)","effort":"high","date":"2026-04-14","stderr":0.0394},"gpqa":{"v":0.7847,"as":"GPT-5.4 nano (high)","effort":"high","date":"2026-04-14","stderr":0.0244},"simpleqa":{"v":0.117,"as":"GPT-5.4 nano (high)","effort":"high","date":"2026-08-27","stderr":0.010169287802713345},"eqLongform":{"v":57.8,"as":"gpt-5.4-nano","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.01,"regions":["us"],"n":4,"tps":80,"ttft":950,"via":"OpenAI","reqs":60878},"open":false,"rankable":true,"scored":7},{"id":"openai/gpt-4.1-nano","name":"OpenAI: GPT-4.1 Nano","org":"OpenAI","slug":"gpt-4.1-nano","matchKey":"gpt41nano","blurb":"For tasks that demand low latency, GPT‑4.1 nano is the fastest and cheapest model in the GPT-4.1 series.","created":1744651369,"cutoff":"2024-06-30","hf":null,"ctx":1047576,"maxOut":32768,"in":["image","text","file"],"out":["text"],"price":{"in":0.1,"out":0.4,"webSearch":0.01,"alt":{"batch":{"in":0.05,"out":0.2}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","maxOut":942818,"in":0.1,"out":0.4,"cacheRead":0.03,"uptime":99.9},{"name":"OpenAI","in":0.1,"out":0.4,"cacheRead":0.025,"uptime":99.6,"tps":41,"ttft":840,"reqs":34912},{"name":"Azure","region":"eu","maxOut":942818,"in":0.11,"out":0.44,"cacheRead":0.033,"uptime":100,"tps":27,"ttft":1919,"reqs":2960}],"scores":{"arenaVision":{"v":1088.8928330187389,"ci":18.3,"rank":124,"votes":1211,"as":"gpt-4.1-nano-2025-04-14","effort":""},"aime":{"v":0.28888888888888886,"as":"GPT-4.1 nano","effort":"","date":"2025-04-14","stderr":0.05970019560922746},"gpqa":{"v":0.48926767676767674,"as":"GPT-4.1 nano","effort":"","date":"2025-04-14","stderr":0.024849734950435368},"simpleqa":{"v":0.06,"as":"GPT-4.1 nano","effort":"","date":"2026-08-31","stderr":0.007513751157474971},"bfcl":{"v":0.33049999999999996,"as":"GPT-4.1-nano-2025-04-14 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":944,"as":"gpt-4.1-nano","version":"Creative Writing v3"},"eqLongform":{"v":32.4,"as":"gpt-4.1-nano","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.03,"regions":["eu"],"n":3,"tps":41,"ttft":840,"via":"OpenAI","reqs":34912},"open":false,"rankable":true,"scored":7},{"id":"openai/gpt-4o","name":"OpenAI: GPT-4o","org":"OpenAI","slug":"gpt-4o","matchKey":"gpt4o","blurb":"GPT-4o (\"o\" for \"omni\") is OpenAI's latest AI model, supporting both text and image inputs with text outputs.","created":1715558400,"cutoff":"2023-10-31","hf":null,"ctx":128000,"maxOut":16384,"in":["text","image","file"],"out":["text"],"price":{"in":2.5,"out":10,"alt":{"batch":{"in":1.25,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","in":2.5,"out":10,"uptime":99.9,"tps":54,"ttft":906,"reqs":4296},{"name":"OpenAI","in":2.5,"out":10,"cacheRead":1.25,"uptime":99.9,"tps":50,"ttft":605,"reqs":7853}],"scores":{"arenaSearch":{"v":1006.180528118307,"ci":11.1,"rank":34,"votes":3411,"as":"api-gpt-4o-search","effort":""},"frontierMath":{"v":0.0035087719298245615,"as":"GPT-4o (Aug 2024)","effort":"","date":"2026-08-27","stderr":0.0035087719298245615},"aime":{"v":0.06388888888888888,"as":"GPT-4o (Aug 2024)","effort":"","date":"2025-02-25","stderr":0.025630380139261548},"gpqa":{"v":0.49210858585858586,"as":"GPT-4o (Aug 2024)","effort":"","date":"2025-01-27","stderr":0.025999270277271707},"simpleqa":{"v":0.26,"as":"GPT-4o (Aug 2024)","effort":"","date":"2026-08-31","stderr":0.01387777332977422},"swebench":{"v":0.30991735537190085,"as":"GPT-4o (Nov 2024)","effort":"","date":"2026-02-11","stderr":0.021042625979133336},"swebenchAgent":{"v":0.38799999999999996,"as":"GPT-4o","agent":"Agentless-1.5","date":"2024-10-28","checked":false}},"perf":{"uptime":99.9,"n":2,"tps":50,"ttft":605,"tpsBest":54,"via":"OpenAI","reqs":7853},"open":false,"rankable":true,"scored":7},{"id":"openai/o1","name":"OpenAI: o1","org":"OpenAI","slug":"o1","matchKey":"o1","blurb":"The latest and strongest model family from OpenAI, o1 is designed to spend more time thinking before responding.","created":1734459999,"cutoff":"2023-10-31","hf":null,"ctx":200000,"maxOut":100000,"in":["text","image","file"],"out":["text"],"price":{"in":15,"out":60,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":15,"out":60,"cacheRead":7.5,"uptime":100,"tps":75,"ttft":9495,"reqs":59}],"scores":{"arenaChat":{"v":1402.129970869491,"ci":4.5,"rank":145,"votes":27807,"as":"o1-2024-12-17","effort":""},"arenaVision":{"v":1193.0058518624535,"ci":10.7,"rank":82,"votes":3694,"as":"o1-2024-12-17","effort":""},"frontierMath":{"v":0.14736842105263157,"as":"o1 (high)","effort":"high","date":"2026-08-28","stderr":0.021034091168852485},"aime":{"v":0.7333333333333333,"as":"o1 (high)","effort":"high","date":"2026-07-15","stderr":0.06666666666666667},"gpqa":{"v":0.7676767676767676,"as":"o1 (high)","effort":"high","date":"2025-02-13","stderr":0.030088629490217445},"simpleqa":{"v":0.411,"as":"o1 (high)","effort":"high","date":"2026-08-31","stderr":0.01556667341859915},"swebenchAgent":{"v":0.6459999999999999,"as":"o1-preview","agent":"W&B Programmer O1 crosscheck5","date":"2025-01-17","checked":false}},"perf":{"uptime":100,"cache":7.5,"n":1,"tps":75,"ttft":9495,"via":"OpenAI","reqs":59},"open":false,"rankable":true,"scored":7},{"id":"deepseek/deepseek-v4-flash-0731","name":"DeepSeek: DeepSeek V4 Flash 0731","org":"DeepSeek","slug":"deepseek-v4-flash-0731","matchKey":"deepseekv4flash0731","blurb":"DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total.","created":1785478908,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V4-Flash-0731","ctx":1310720,"maxOut":943718,"in":["text"],"out":["text"],"price":{"in":0.04,"out":0.64,"alt":{"batch":{"in":0.11,"out":0.33}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"OpenInference","quant":"fp8","ctx":1048576,"in":0.03,"out":0.8,"cacheRead":0.008,"uptime":99.5,"tps":22,"ttft":1326,"reqs":82438},{"name":"Sail Research","quant":"fp4","ctx":1048576,"in":0.038,"out":0.55,"cacheRead":0.0228,"uptime":99.9,"tps":42,"ttft":869,"reqs":27139},{"name":"Relace","quant":"fp4","ctx":1048576,"in":0.04,"out":0.64,"cacheRead":0.016,"uptime":99.8,"noSchema":1,"tps":30,"ttft":923,"reqs":203236},{"name":"StreamLake","quant":"fp8","ctx":1024000,"maxOut":384000,"in":0.0528,"out":0.1584,"cacheRead":0.00168,"uptime":98.6,"noSchema":1,"tps":43,"ttft":1569,"reqs":80820},{"name":"DeepInfra","quant":"fp8","ctx":1048576,"maxOut":384000,"in":0.06,"out":0.18,"cacheRead":0.015,"uptime":99.6,"tps":23,"ttft":816,"reqs":238650},{"name":"Inceptron","quant":"fp4","ctx":1048576,"in":0.06,"out":0.3,"cacheRead":0.04,"uptime":99.6,"noSchema":1,"tps":44,"ttft":646,"reqs":13003,"region":"eu"},{"name":"Reka","quant":"fp4","ctx":262144,"maxOut":131072,"in":0.088,"out":0.528,"cacheRead":0.0056,"uptime":98.8,"tps":87,"ttft":735,"reqs":11782},{"name":"Makora","ctx":1000000,"maxOut":384000,"in":0.09,"out":0.195,"cacheRead":0.0196,"uptime":97.4,"tps":73,"ttft":486,"reqs":13375},{"name":"Wafer","ctx":1048576,"in":0.1,"out":0.25,"cacheRead":0.05,"uptime":100,"tps":101,"ttft":338,"reqs":108798},{"name":"DigitalOcean","ctx":1048576,"in":0.119,"out":0.238,"cacheRead":0.0238,"uptime":99.9,"noSchema":1,"tps":16,"ttft":379,"reqs":59312},{"name":"BaseTen","quant":"fp8","ctx":1048576,"maxOut":384000,"in":0.13,"out":0.26,"cacheRead":0.028,"uptime":100,"noSchema":1},{"name":"BaseTen","quant":"fp8","ctx":1048576,"maxOut":384000,"in":0.13,"out":0.26,"cacheRead":0.028,"uptime":100,"noSchema":1,"tps":98,"ttft":386,"reqs":3993,"region":"us"},{"name":"CoreWeave","quant":"fp8","ctx":262144,"maxOut":235929,"in":0.13,"out":0.28,"cacheRead":0.07,"uptime":100,"noSchema":1,"tps":101,"ttft":386,"reqs":235623},{"name":"Nebius","quant":"fp8","ctx":1024000,"maxOut":921600,"in":0.14,"out":0.28,"uptime":62.7,"tps":54,"ttft":1072,"reqs":8184},{"name":"Together","ctx":1048576,"in":0.14,"out":0.28,"cacheRead":0.03,"uptime":99.6,"tps":34,"ttft":949,"reqs":23091},{"name":"Parasail","quant":"fp8","ctx":1048576,"in":0.14,"out":0.28,"cacheRead":0.05,"uptime":99.5,"tps":42,"ttft":678,"reqs":25580},{"name":"Morph","ctx":1048576,"in":0.141953,"out":0.399625,"cacheRead":0.035938,"uptime":94.8,"tps":18,"ttft":1539,"reqs":2899},{"name":"Venice","ctx":1000000,"maxOut":32768,"in":0.175,"out":0.35,"cacheRead":0.035,"uptime":99.1,"tps":40,"ttft":985,"reqs":8264},{"name":"Mancer 2","quant":"fp8","ctx":1048576,"in":0.2,"out":0.6,"uptime":99.1,"tps":32,"ttft":1061,"reqs":1453},{"name":"Fireworks","ctx":1048576,"in":0.22,"out":0.66,"cacheRead":0.007,"uptime":98.3,"tps":118,"ttft":735,"reqs":7225},{"name":"SiliconFlow","quant":"fp8","ctx":1048576,"maxOut":393216,"in":0.22,"out":0.66,"cacheRead":0.028,"uptime":98.2,"tps":52,"ttft":1642,"reqs":5023},{"name":"GMICloud","quant":"fp8","ctx":1048575,"maxOut":943717,"in":0.286,"out":0.858,"cacheRead":0.0091,"uptime":99.9,"noSchema":1,"tps":54,"ttft":1900,"reqs":16213},{"name":"NextBit","quant":"fp8","ctx":1048576,"in":0.352,"out":1.056,"cacheRead":0.012,"uptime":98.8,"tps":45,"ttft":2159,"reqs":896},{"name":"Alibaba","ctx":1000000,"maxOut":393216,"in":0.352,"out":1.056,"cacheRead":0.0352,"uptime":100,"tps":38,"ttft":902,"reqs":6176},{"name":"Novita","quant":"fp8","ctx":1048576,"maxOut":393216,"in":0.4092,"out":1.2276,"cacheRead":0.02604,"uptime":100,"noSchema":1,"tps":35,"ttft":921,"reqs":271087},{"name":"Phala","ctx":1048576,"maxOut":393216,"in":0.44,"out":1.32,"cacheRead":0.028,"uptime":99.7,"tps":55,"ttft":2307,"reqs":11200},{"name":"Baidu","quant":"fp8","ctx":1048576,"maxOut":131072,"in":0.44,"out":1.32,"cacheRead":0.014,"uptime":99.5,"tps":51,"ttft":1037,"reqs":2829},{"name":"AtlasCloud","quant":"fp4","ctx":1048576,"maxOut":393216,"in":0.44,"out":1.32,"cacheRead":0.028,"uptime":99.8,"tps":58,"ttft":1757,"reqs":1443},{"name":"Cloudflare","maxOut":1179648,"in":0.44,"out":1.32,"cacheRead":0.014,"uptime":100,"tps":30,"ttft":857,"reqs":8377}],"scores":{"frontierMath":{"v":0.5754385964912281,"as":"DeepSeek V4 Flash 0731 (max)","effort":"max","date":"2026-08-02","stderr":0.029329899789943818},"aime":{"v":0.9444444444444444,"as":"DeepSeek V4 Flash 0731 (max)","effort":"max","date":"2026-08-02","stderr":0.020854366486931897},"gpqa":{"v":0.9103535353535354,"as":"DeepSeek V4 Flash 0731 (max)","effort":"max","date":"2026-08-02","stderr":0.017655466909172534},"simpleqa":{"v":0.33633633633633636,"as":"DeepSeek V4 Flash 0731 (max)","effort":"max","date":"2026-08-27","stderr":0.01495531772489948},"eqWriting":{"v":1438.4,"as":"deepseek-ai/DeepSeek-V4-Flash-0731","version":"Creative Writing v3"},"eqLongform":{"v":61.2,"as":"deepseek-ai/DeepSeek-V4-Flash-0731","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.008,"n":29,"tps":35,"ttft":921,"tpsBest":118,"via":"Novita","reqs":271087,"regions":["eu","us"]},"open":true,"rankable":true,"scored":6},{"id":"qwen/qwen3.8-27b","name":"Qwen: Qwen3.8 27B","org":"Qwen","slug":"qwen3.8-27b","matchKey":"qwen3827b","blurb":"Qwen3.8 27B is an open-weight dense vision-language model from Qwen.","created":1786722910,"cutoff":null,"hf":"Qwen/Qwen3.8-27B","ctx":1000000,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":0.42,"out":3,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","medium","low"]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"fp4","ctx":262144,"maxOut":32768,"in":0.1,"out":1.8,"uptime":98.7,"tps":1,"ttft":2042,"reqs":3435},{"name":"DekaLLM","ctx":262144,"maxOut":235929,"in":0.12,"out":2.5,"cacheRead":0.05,"uptime":99.8,"tps":41,"ttft":2554,"reqs":6736},{"name":"Reka","quant":"fp8","ctx":262144,"in":0.12,"out":2.5,"cacheRead":0.05,"uptime":100,"tps":54,"ttft":818,"reqs":2061},{"name":"DeepInfra","quant":"bf16","ctx":262144,"maxOut":235929,"in":0.15,"out":1.875,"cacheRead":0.0375,"uptime":95.5,"tps":36,"ttft":802,"reqs":1914},{"name":"Phala","ctx":262144,"maxOut":235929,"in":0.1992,"out":2.075,"cacheRead":0.0415,"uptime":96.1,"tps":61,"ttft":522,"reqs":1500},{"name":"Wafer","ctx":262144,"maxOut":235929,"in":0.2,"out":2.5,"cacheRead":0.05,"uptime":99.6,"tps":71,"ttft":377,"reqs":970},{"name":"Mancer 2","quant":"fp8","ctx":262144,"maxOut":235929,"in":0.2,"out":2.5,"uptime":98.3,"tps":68,"ttft":778,"reqs":236},{"name":"Parasail","quant":"fp8","ctx":262144,"maxOut":235929,"in":0.24,"out":2.2,"cacheRead":0.05,"uptime":99.8,"tps":52,"ttft":634,"reqs":7995},{"name":"Chutes","quant":"fp8","ctx":262144,"maxOut":65536,"in":0.24,"out":2.2,"cacheRead":0.024,"uptime":99,"tps":30,"ttft":2192,"reqs":7018},{"name":"AkashML","quant":"fp8","ctx":262144,"in":0.25,"out":2.2,"cacheRead":0.05,"uptime":100,"tps":70,"ttft":431,"reqs":3281},{"name":"Ionstream","quant":"fp8","ctx":262144,"maxOut":65536,"in":0.28,"out":2.55,"cacheRead":0.1,"uptime":99.6,"tps":70,"ttft":412,"reqs":363},{"name":"Io Net","quant":"fp8","ctx":65536,"maxOut":58982,"in":0.285,"out":2.66,"cacheRead":0.171,"uptime":99.8,"tps":54,"ttft":667,"reqs":155},{"name":"CoreWeave","quant":"fp8","ctx":262144,"maxOut":235929,"in":0.4,"out":3,"cacheRead":0.15,"uptime":99.6,"tps":58,"ttft":481,"reqs":462},{"name":"Novita","in":0.42,"out":3,"cacheRead":0.085,"uptime":99.9,"noSchema":1,"tps":36,"ttft":2874,"reqs":3141},{"name":"Alibaba","in":0.425,"out":2.55,"cacheRead":0.085,"cacheWrite":0.53125,"uptime":100,"noSchema":1,"tps":41,"ttft":1595,"reqs":530},{"name":"Cloudflare","ctx":262144,"maxOut":235929,"in":0.45,"out":3.2,"cacheRead":0.05,"uptime":87.3,"tps":31,"ttft":670,"reqs":34},{"name":"Venice","quant":"fp8","ctx":262144,"maxOut":65536,"in":0.45,"out":3.2,"uptime":97.6,"tps":89,"ttft":875,"reqs":57}],"scores":{"arenaChat":{"v":1437.0954739209874,"ci":6.5,"rank":90,"votes":10697,"as":"qwen3.8-27b","effort":""},"arenaWebdev":{"v":1593.2860031272262,"ci":9.5,"rank":18,"votes":4913,"as":"qwen3.8-27b","effort":""},"arenaVision":{"v":1244.1862130800462,"ci":9.9,"rank":49,"votes":4577,"as":"qwen3.8-27b","effort":""},"arenaImageCode":{"v":1582.3712647052985,"ci":12.7,"rank":11,"votes":2439,"as":"qwen3.8-27b","effort":""},"eqWriting":{"v":1668.4,"as":"Qwen/Qwen3.8-27B","version":"Creative Writing v3"},"eqLongform":{"v":53.3,"as":"Qwen/Qwen3.8-27B","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":17,"tps":52,"ttft":634,"tpsBest":89,"via":"Parasail","reqs":7995},"open":true,"rankable":true,"scored":6},{"id":"moonshotai/kimi-k2.7-code","name":"MoonshotAI: Kimi K2.7 Code","org":"MoonshotAI","slug":"kimi-k2.7-code","matchKey":"kimik27code","blurb":"MoonshotAI: Kimi K2.7 Code is a coding-focused model in Moonshot AI's Kimi K2 family, built to complete end-to-end programming tasks reliably over long contexts.","created":1781266361,"cutoff":null,"hf":"moonshotai/Kimi-K2.7-Code","ctx":262144,"maxOut":235929,"in":["text","image"],"out":["text"],"price":{"in":0.7062,"out":3.21},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","maxOut":16384,"in":0.68,"out":3.4,"cacheRead":0.136,"noSchema":1},{"name":"Inceptron","quant":"int4","in":0.7062,"out":3.21,"cacheRead":0.18,"uptime":99.8,"tps":60,"ttft":939,"reqs":1738,"region":"eu"},{"name":"CoreWeave","quant":"int4","in":0.71,"out":3.5,"cacheRead":0.15,"uptime":100,"tps":102,"ttft":1294,"reqs":301},{"name":"StreamLake","ctx":256000,"maxOut":32000,"in":0.7125,"out":3,"cacheRead":0.1425,"uptime":99.8,"tps":27,"ttft":1676,"reqs":103},{"name":"Venice","quant":"int4","ctx":256000,"maxOut":65536,"in":0.75,"out":3.5,"cacheRead":0.16,"uptime":96.5,"tps":48,"ttft":1292,"reqs":162},{"name":"ModelRun","quant":"fp4","in":0.85,"out":3.75,"cacheRead":0.16,"uptime":100,"tps":138,"ttft":540,"reqs":461},{"name":"SiliconFlow","quant":"fp8","in":0.85916,"out":3.8,"cacheRead":0.17993,"uptime":99.9,"noSchema":1,"tps":20,"ttft":2743,"reqs":147},{"name":"Novita","quant":"int4","in":0.912,"out":3.84,"cacheRead":0.1824,"uptime":100,"tps":26,"ttft":2182,"reqs":204},{"name":"Fireworks","in":0.95,"out":4,"cacheRead":0.19,"uptime":95.4,"tps":78,"ttft":814,"reqs":50},{"name":"BaseTen","quant":"fp4","ctx":262000,"maxOut":235800,"in":0.95,"out":4,"cacheRead":0.16,"uptime":100,"noSchema":1,"tps":132,"ttft":452,"reqs":104,"region":"us"},{"name":"GMICloud","quant":"fp8","in":0.95,"out":4,"cacheRead":0.19,"uptime":99.6,"tps":28,"ttft":3959,"reqs":24},{"name":"Alibaba","quant":"fp8","maxOut":16384,"in":0.95,"out":4,"cacheRead":0.19,"uptime":99.9,"tps":29,"ttft":5170,"reqs":30},{"name":"Cloudflare","in":0.95,"out":4,"cacheRead":0.19,"uptime":100,"tps":94,"ttft":609,"reqs":62},{"name":"Moonshot AI","quant":"int4","in":0.95,"out":4,"cacheRead":0.19,"uptime":100},{"name":"Moonshot AI","quant":"int4","in":1.9,"out":8,"cacheRead":0.38,"uptime":99.9,"tps":27,"ttft":1181,"reqs":106}],"scores":{"arenaWebdev":{"v":1472.7184462175373,"ci":9.7,"rank":51,"votes":4683,"as":"kimi-k2.7-code","effort":""},"arenaImageCode":{"v":1510.8299106941445,"ci":21,"rank":26,"votes":845,"as":"kimi-k2.7-code","effort":""},"frontierMath":{"v":0.5403508771929825,"as":"Kimi K2.7 Code","effort":"","date":"2026-06-13","stderr":0.02957276813514758},"aime":{"v":0.9555555555555556,"as":"Kimi K2.7 Code","effort":"","date":"2026-08-07","stderr":0.031067790907534736},"gpqa":{"v":0.8787878787878788,"as":"Kimi K2.7 Code","effort":"","date":"2026-08-07","stderr":0.02325315795194205},"simpleqa":{"v":0.365,"as":"Kimi K2.7 Code","effort":"","date":"2026-08-27","stderr":0.015231776226264846}},"perf":{"uptime":100,"cache":0.136,"n":15,"tps":60,"ttft":939,"tpsBest":138,"via":"Inceptron","reqs":1738,"regions":["eu","us"]},"open":true,"rankable":true,"scored":6},{"id":"minimax/minimax-m3","name":"MiniMax: MiniMax M3","org":"MiniMax","slug":"minimax-m3","matchKey":"minimaxm3","blurb":"MiniMax-M3 is a multimodal foundation model from MiniMax.","created":1780245374,"cutoff":null,"hf":"MiniMaxAI/Minimax-M3","ctx":1048576,"maxOut":512000,"in":["text","image","video"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"CoreWeave","quant":"fp4","ctx":262144,"maxOut":235929,"in":0.23,"out":0.96,"cacheRead":0.05,"uptime":99.7,"tps":25,"ttft":1260,"reqs":4830},{"name":"GMICloud","quant":"fp8","maxOut":524288,"in":0.24,"out":0.96,"cacheRead":0.048,"uptime":99.8,"noSchema":1,"tps":45,"ttft":1531,"reqs":2297},{"name":"DeepInfra","quant":"fp8","ctx":524288,"in":0.28,"out":1.1,"cacheRead":0.056,"uptime":98.6,"noSchema":1,"tps":19,"ttft":1488,"reqs":233},{"name":"StreamLake","quant":"fp8","ctx":1000000,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":99.2,"noTools":1,"noSchema":1,"tps":64,"ttft":1655,"reqs":535},{"name":"Venice","quant":"fp8","ctx":524288,"maxOut":65536,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":91.6,"noSchema":1,"tps":22,"ttft":1914,"reqs":2886},{"name":"Together","ctx":524288,"maxOut":471859,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":99.8,"tps":84,"ttft":696,"reqs":2255},{"name":"Parasail","quant":"fp8","maxOut":524288,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":98.4,"noSchema":1,"tps":129,"ttft":714,"reqs":471},{"name":"AtlasCloud","quant":"fp8","ctx":524300,"maxOut":524288,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":94.8,"noSchema":1,"tps":82,"ttft":2308,"reqs":2492},{"name":"Novita","quant":"fp8","ctx":1000000,"maxOut":131072,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":99.9,"noSchema":1,"tps":62,"ttft":1669,"reqs":1019},{"name":"Minimax","quant":"fp8","ctx":524288,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":98.5,"noSchema":1,"tps":67,"ttft":1116,"reqs":44199},{"name":"Mara","maxOut":943718,"in":0.6,"out":2.4,"uptime":35.9,"noSchema":1,"tps":94,"ttft":1745,"reqs":66},{"name":"SambaNova","maxOut":943718,"in":0.6,"out":2.4,"uptime":99.2,"noSchema":1,"tps":100,"ttft":1566,"reqs":78},{"name":"ModelRun","quant":"fp4","maxOut":943718,"in":0.75,"out":3,"cacheRead":0.15,"uptime":93.5,"tps":134,"ttft":553,"reqs":462}],"scores":{"arenaChat":{"v":1441.2503431504656,"ci":4.2,"rank":84,"votes":48540,"as":"minimax-m3","effort":""},"arenaWebdev":{"v":1486.7017314128002,"ci":6.4,"rank":48,"votes":12686,"as":"minimax-m3","effort":""},"arenaVision":{"v":1236.5628547782298,"ci":7.1,"rank":56,"votes":15192,"as":"minimax-m3","effort":""},"arenaDocument":{"v":1434.7938037275746,"ci":8.3,"rank":32,"votes":6314,"as":"minimax-m3","effort":""},"arenaImageCode":{"v":1472.6756556921805,"ci":10.5,"rank":33,"votes":3645,"as":"minimax-m3","effort":""},"eqChat":{"v":1149.7,"as":"minimax/minimax-m3","version":"EQ-Bench 4","ciLow":1135.7,"ciHigh":1161.9,"boardAsOf":"2026-07-26T01:26:10+00:00"}},"perf":{"uptime":99.9,"cache":0.05,"n":13,"tps":67,"ttft":1116,"tpsBest":134,"via":"Minimax","reqs":44199},"open":true,"rankable":true,"scored":6},{"id":"meta-llama/llama-4-maverick","name":"Meta: Llama 4 Maverick","org":"Meta","slug":"llama-4-maverick","matchKey":"llama4maverick","blurb":"Llama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active parameters per forward...","created":1743881822,"cutoff":"2024-08-31","hf":"meta-llama/Llama-4-Maverick-17B-128E-Instruct","ctx":1048576,"maxOut":16384,"in":["text","image"],"out":["text"],"price":{"in":0.1875,"out":0.6525},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DigitalOcean","ctx":128000,"in":0.1875,"out":0.6525,"uptime":99.9,"tps":8,"ttft":718,"reqs":25332},{"name":"DeepInfra","quant":"fp8","in":0.2,"out":0.8,"uptime":99.7,"noTools":1,"tps":36,"ttft":411,"reqs":15297},{"name":"Novita","quant":"fp8","maxOut":8192,"in":0.27,"out":0.85,"uptime":99.6,"noTools":1,"tps":27,"ttft":595,"reqs":1704},{"name":"Parasail","quant":"fp8","ctx":524288,"maxOut":32768,"in":0.35,"out":1,"cacheRead":0.17,"uptime":99.9,"tps":35,"ttft":435,"reqs":5918},{"name":"Google","region":"us","ctx":524288,"maxOut":8192,"in":0.35,"out":1.15}],"scores":{"aime":{"v":0.20555555555555555,"as":"Llama 4 Maverick (FP8)","effort":"","date":"2025-04-08","stderr":0.04948528559541873},"gpqa":{"v":0.6698232323232324,"as":"Llama 4 Maverick (FP8)","effort":"","date":"2025-04-08","stderr":0.028244823529728173},"swebenchAgent":{"v":0.2104,"as":"Llama 4 Maverick Instruct","agent":"mini-SWE-agent","date":"2025-07-20","checked":true},"bfcl":{"v":0.3729,"as":"Llama-4-Maverick-17B-128E-Instruct-FP8 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":858.4,"as":"meta-llama/Llama-4-Maverick-17B-128E-Instruct","version":"Creative Writing v3"},"eqLongform":{"v":31.9,"as":"meta-llama/Llama-4-Maverick-17B-128E-Instruct","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"regions":["us"],"n":5,"tps":8,"ttft":718,"tpsBest":36,"via":"DigitalOcean","reqs":25332},"open":true,"prose":{"slop":23.733409337760076,"trigram":1.1246600984462742,"notXButY":0.1300421672616044,"repetition":13.598528842431282,"mattr":0.46189347637336087,"grade":11.038905174629242,"sentence":18.748843498683367,"words":259634,"samples":300,"as":"meta-llama/llama-4-maverick"},"rankable":true,"scored":6},{"id":"x-ai/grok-4.3","name":"SpaceXAI: Grok 4.3","org":"SpaceXAI","slug":"grok-4.3","matchKey":"grok43","blurb":"Grok 4.3 is a reasoning model from SpaceXAI.","created":1777591821,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":900000,"in":["text","image","file"],"out":["text"],"price":{"in":1.25,"out":2.5,"webSearch":0.005,"alt":{"batch":{"in":1,"out":2}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","none"]},"moderated":false,"providers":[{"name":"xAI","in":1.25,"out":2.5,"cacheRead":0.2,"uptime":99.5},{"name":"xAI","in":1.25,"out":2.5,"cacheRead":0.2,"uptime":99.6},{"name":"xAI","in":2.5,"out":5,"cacheRead":0.4,"uptime":98.2},{"name":"xAI","in":2.5,"out":5,"cacheRead":0.4,"uptime":93.7,"tps":75,"ttft":665,"reqs":33623}],"scores":{"arenaChat":{"v":1442.6505317838596,"ci":3.7,"rank":81,"votes":66801,"as":"grok-4.3","effort":""},"arenaWebdev":{"v":1357.1677364173438,"ci":6.4,"rank":96,"votes":14192,"as":"grok-4.3","effort":""},"arenaVision":{"v":1243.175226505605,"ci":6.3,"rank":51,"votes":22688,"as":"grok-4.3","effort":""},"arenaImageCode":{"v":1373.4739143958518,"ci":9.9,"rank":46,"votes":5444,"as":"grok-4.3","effort":""},"arenaSearch":{"v":1164.8004512187958,"ci":5.2,"rank":23,"votes":91483,"as":"grok-4.3","effort":""},"eqChat":{"v":1074.8,"as":"x-ai/grok-4.3","version":"EQ-Bench 4","ciLow":1058,"ciHigh":1089.7,"boardAsOf":"2026-07-26T01:26:10+00:00"}},"perf":{"uptime":99.6,"cache":0.2,"n":4,"tps":75,"ttft":665,"via":"xAI","reqs":33623},"open":false,"rankable":true,"scored":6},{"id":"z-ai/glm-4.6","name":"Z.ai: GLM 4.6","org":"Z.ai","slug":"glm-4.6","matchKey":"glm46","blurb":"Compared with GLM-4.5, this generation brings several key improvements: Longer context window: The context window has been expanded from 128K to 200K tokens, enabling the model to handle more complex...","created":1759235576,"cutoff":"2025-03-31","hf":"zai-org/GLM-4.6","ctx":204800,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.43,"out":1.75},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Venice","quant":"fp4","ctx":198000,"in":0.43,"out":1.75,"cacheRead":0.08,"uptime":99.7,"tps":36,"ttft":780,"reqs":4183},{"name":"DeepInfra","quant":"fp4","ctx":202752,"maxOut":131072,"in":0.5,"out":2,"cacheRead":0.1,"uptime":99.1,"noSchema":1,"tps":24,"ttft":517,"reqs":115},{"name":"Novita","quant":"bf16","maxOut":131072,"in":0.55,"out":2.2,"cacheRead":0.11,"uptime":99.6,"noSchema":1,"tps":25,"ttft":3490,"reqs":13126},{"name":"Z.AI","quant":"fp4","ctx":202752,"maxOut":131072,"in":0.6,"out":2.2,"cacheRead":0.11,"uptime":92.5,"noSchema":1,"tps":25,"ttft":9152,"reqs":984}],"scores":{"arenaChat":{"v":1424.7439615431895,"ci":3.9,"rank":112,"votes":35061,"as":"glm-4.6","effort":""},"arenaWebdev":{"v":1340.6016580660078,"ci":10.9,"rank":100,"votes":6600,"as":"glm-4.6","effort":""},"swebenchAgent":{"v":0.682,"as":"GLM-4.6","agent":"Undisclosed","date":"2025-09-30","checked":false},"bfcl":{"v":0.7238,"as":"GLM-4.6 (FC thinking)","effort":"default","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1408.6,"as":"zai-org/GLM-4.6","version":"Creative Writing v3"},"eqLongform":{"v":57.3,"as":"zai-org/GLM-4.6","version":"Longform Writing v1.11"}},"perf":{"uptime":99.7,"cache":0.08,"n":4,"tps":25,"ttft":3490,"tpsBest":36,"via":"Novita","reqs":13126},"open":true,"rankable":true,"scored":6},{"id":"meta/muse-spark-1.2","name":"Meta: Muse Spark 1.2","org":"Meta","slug":"muse-spark-1.2","matchKey":"musespark12","blurb":"Muse Spark 1.2 is a reasoning model from Meta, designed for complex agentic tasks.","created":1785959287,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":943718,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":1.25,"out":4.25,"webSearch":0.0025},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"Meta","in":1.25,"out":4.25,"cacheRead":0.15,"uptime":100,"tps":125,"ttft":3466,"reqs":290}],"scores":{"arenaChat":{"v":1499.5893180176977,"ci":10.5,"rank":4,"votes":3227,"as":"muse-spark-1.2 (xHigh)","effort":"xhigh"},"arenaWebdev":{"v":1534.3767704428442,"ci":13.9,"rank":34,"votes":2065,"as":"muse-spark-1.2 (xHigh)","effort":"xhigh"},"arenaVision":{"v":1291.5423883838137,"ci":14.8,"rank":9,"votes":1841,"as":"muse-spark-1.2 (xHigh)","effort":"xhigh"},"simpleqa":{"v":0.6030150753768844,"as":"Muse Spark 1.2 (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015518791563606454},"eqWriting":{"v":1835.2,"as":"muse-spark-1.2","version":"Creative Writing v3"},"eqLongform":{"v":81.5,"as":"muse-spark-1.2","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.15,"n":1,"tps":125,"ttft":3466,"via":"Meta","reqs":290},"open":false,"rankable":true,"scored":6},{"id":"qwen/qwen3.6-max-preview","name":"Qwen: Qwen3.6 Max Preview","org":"Qwen","slug":"qwen3.6-max-preview","matchKey":"qwen36maxpreview","blurb":"Qwen3.6-Max-Preview is a proprietary frontier model from Alibaba Cloud built on a sparse mixture-of-experts architecture with approximately 1 trillion total parameters.","created":1777260242,"cutoff":null,"hf":null,"ctx":262144,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":1.027,"out":6.162},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":1.027,"out":6.162,"cacheWrite":1.28375,"uptime":100,"tps":38,"ttft":1642,"reqs":85}],"scores":{"arenaChat":{"v":1460.00905823416,"ci":8.4,"rank":53,"votes":5186,"as":"qwen3.6-max-preview","effort":"max"},"arenaWebdev":{"v":1479.0595857062222,"ci":12.8,"rank":49,"votes":2601,"as":"qwen3.6-max-preview","effort":"max"},"aime":{"v":0.9111111111111111,"as":"Qwen 3.6 Max (Preview)","effort":"max","date":"2026-08-07","stderr":0.04290254662948546},"gpqa":{"v":0.8737373737373737,"as":"Qwen 3.6 Max (Preview)","effort":"max","date":"2026-08-07","stderr":0.023664359402880177},"simpleqa":{"v":0.5200400801603207,"as":"Qwen 3.6 Max (Preview)","effort":"max","date":"2026-08-27","stderr":0.015822434816352416},"swebench":{"v":0.7665289256198347,"as":"Qwen 3.6 Max (Preview)","effort":"max","date":"2026-05-28","stderr":0.01924895909142592}},"perf":{"uptime":100,"n":1,"tps":38,"ttft":1642,"via":"Alibaba","reqs":85},"open":false,"rankable":true,"scored":6},{"id":"qwen/qwen3.7-plus","name":"Qwen: Qwen3.7 Plus","org":"Qwen","slug":"qwen3.7-plus","matchKey":"qwen37plus","blurb":"Qwen3.7-Plus is a cost-effective model in Alibaba's Qwen3.7 series.","created":1780491783,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":131072,"in":["text","image"],"out":["text"],"price":{"in":0.32,"out":1.28},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.32,"out":1.28,"cacheRead":0.064,"cacheWrite":0.4,"uptime":100,"tps":12,"ttft":1066,"reqs":35205}],"scores":{"arenaChat":{"v":1455.5650312094194,"ci":4.5,"rank":62,"votes":39334,"as":"qwen3.7-plus","effort":""},"arenaVision":{"v":1262.7374762273182,"ci":7.6,"rank":36,"votes":10656,"as":"qwen3.7-plus","effort":""},"arenaDocument":{"v":1444.1444013632026,"ci":9,"rank":30,"votes":4591,"as":"qwen3.7-plus","effort":""},"frontierMath":{"v":0.34385964912280703,"as":"Qwen3.7 Plus","effort":"","date":"2026-08-29","stderr":0.028185763989072583},"aime":{"v":0.9333333333333333,"as":"Qwen3.7 Plus","effort":"","date":"2026-08-07","stderr":0.03760507165451775},"gpqa":{"v":0.8787878787878788,"as":"Qwen3.7 Plus","effort":"","date":"2026-08-07","stderr":0.02325315795194205}},"perf":{"uptime":100,"cache":0.064,"n":1,"tps":12,"ttft":1066,"via":"Alibaba","reqs":35205},"open":false,"rankable":true,"scored":6},{"id":"minimax/minimax-m2.5","name":"MiniMax: MiniMax M2.5","org":"MiniMax","slug":"minimax-m2.5","matchKey":"minimaxm25","blurb":"MiniMax-M2.5 is a SOTA large language model designed for real-world productivity.","created":1770908502,"cutoff":null,"hf":"MiniMaxAI/MiniMax-M2.5","ctx":204800,"maxOut":128000,"in":["text"],"out":["text"],"price":{"in":0.27,"out":1.08},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Venice","ctx":198000,"maxOut":32768,"in":0.27,"out":0.95,"cacheRead":0.03,"uptime":43.6,"noSchema":1,"tps":9,"ttft":2654,"reqs":19},{"name":"StreamLake","ctx":200000,"in":0.27,"out":1.08,"cacheRead":0.027,"uptime":99.8,"tps":56,"ttft":1353,"reqs":970},{"name":"AtlasCloud","quant":"fp8","ctx":196608,"maxOut":176947,"in":0.295,"out":1.2,"cacheRead":0.06,"uptime":99.6,"tps":37,"ttft":1915,"reqs":100},{"name":"DigitalOcean","ctx":65536,"maxOut":58982,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":100,"tps":56,"ttft":495,"reqs":76},{"name":"Friendli","ctx":196608,"maxOut":176947,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":100,"tps":148,"ttft":516,"reqs":340},{"name":"Novita","quant":"fp8","maxOut":131100,"in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.8,"noSchema":1,"tps":10,"ttft":1197,"reqs":42},{"name":"Minimax","quant":"fp8","maxOut":131072,"in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.6,"noSchema":1},{"name":"Minimax","quant":"fp8","maxOut":131072,"in":0.6,"out":2.4,"cacheRead":0.06,"uptime":99.3,"noSchema":1,"tps":48,"ttft":1742,"reqs":145}],"scores":{"arenaChat":{"v":1390.3222013080435,"ci":4,"rank":162,"votes":40843,"as":"minimax-m2.5","effort":""},"arenaWebdev":{"v":1384.1303886798787,"ci":8.5,"rank":85,"votes":7186,"as":"minimax-m2.5","effort":""},"swebenchAgent":{"v":0.758,"as":"MiniMax M2.5","agent":"mini-SWE-agent","date":"2026-02-17","checked":false},"eqWriting":{"v":1358.4,"as":"minimax/minimax-m2.5","version":"Creative Writing v3"},"eqLongform":{"v":65.3,"as":"minimax/minimax-m2.5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.03,"n":8,"tps":56,"ttft":1353,"tpsBest":148,"via":"StreamLake","reqs":970},"open":true,"rankable":true,"scored":5},{"id":"google/gemini-2.5-flash","name":"Google: Gemini 2.5 Flash","org":"Google","slug":"gemini-2.5-flash","matchKey":"gemini25flash","blurb":"Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks.","created":1750172488,"cutoff":"2025-01-31","hf":null,"ctx":1048576,"maxOut":65535,"in":["file","image","text","audio","video"],"out":["text"],"price":{"in":0.3,"out":2.5,"imageIn":3e-7,"audioIn":1,"webSearch":0.014,"alt":{"batch":{"in":0.15,"out":1.25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Google","region":"eu","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":99},{"name":"Google","region":"global","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":99.5},{"name":"Google","in":0.54,"out":4.5,"cacheRead":0.054,"cacheWrite":0.15,"uptime":99.9},{"name":"Google AI Studio","in":0.15,"out":1.25,"cacheRead":0.015,"cacheWrite":0.041667,"uptime":100},{"name":"Google AI Studio","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":85.8,"tps":116,"ttft":1502,"reqs":294,"region":"eu"},{"name":"Google AI Studio","in":0.54,"out":4.5,"cacheRead":0.054,"cacheWrite":0.15,"uptime":99.9,"tps":55,"ttft":998,"reqs":10058}],"scores":{"arenaChat":{"v":1409.6933534261525,"ci":2.5,"rank":138,"votes":122732,"as":"gemini-2.5-flash","effort":""},"arenaVision":{"v":1215.0379990666236,"ci":4.7,"rank":69,"votes":58893,"as":"gemini-2.5-flash","effort":""},"aime":{"v":0.7305555555555555,"as":"Gemini 2.5 Flash Preview (Apr 2025)","effort":"","date":"2025-04-22","stderr":0.05532781604070319},"swebenchAgent":{"v":0.2873,"as":"Gemini 2.5 Flash","agent":"mini-SWE-agent","date":"2025-07-26","checked":true},"bfcl":{"v":0.5624,"as":"Gemini-2.5-Flash (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":100,"cache":0.015,"regions":["eu","global"],"n":7,"tps":89,"ttft":691,"tpsBest":116,"via":"Google","reqs":201221},"open":false,"prose":{"slop":40.15792442188381,"trigram":0.9512050121187177,"notXButY":0.3864697004930431,"repetition":7.3942314652988195,"mattr":0.5720884136472598,"grade":10.773618218901682,"sentence":15.75233578373251,"words":328005,"samples":300,"as":"google/gemini-2.5-flash"},"rankable":true,"scored":5},{"id":"xiaomi/mimo-v2.5-pro","name":"Xiaomi: MiMo-V2.5-Pro","org":"Xiaomi","slug":"mimo-v2.5-pro","matchKey":"mimov25pro","blurb":"MiMo-V2.5-Pro is Xiaomi’s flagship model, delivering strong performance in general agentic capabilities, complex software engineering, and long-horizon tasks, with top rankings on benchmarks such as ClawEval, GDPVal, and","created":1776874273,"cutoff":null,"hf":"XiaomiMiMo/MiMo-V2.5-Pro","ctx":1050000,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.435,"out":0.87},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"GMICloud","quant":"bf16","maxOut":945000,"in":0.3045,"out":0.609,"cacheRead":0.0028,"uptime":58.6,"noTools":1,"noSchema":1,"tps":25,"ttft":4618,"reqs":88},{"name":"DeepInfra","quant":"fp8","ctx":1048576,"in":0.39,"out":1.17,"cacheRead":0.078,"uptime":91.3,"tps":28,"ttft":1332,"reqs":595},{"name":"AtlasCloud","quant":"fp8","ctx":1024000,"in":0.435,"out":0.87,"cacheRead":0.0036,"uptime":95.2,"noSchema":1,"tps":26,"ttft":5808,"reqs":491},{"name":"Xiaomi","quant":"fp8","ctx":1048576,"in":0.435,"out":0.87,"cacheRead":0.0036,"uptime":99.2,"noSchema":1,"tps":25,"ttft":3419,"reqs":13422},{"name":"DigitalOcean","ctx":262144,"maxOut":235929,"in":0.48,"out":1.8,"cacheRead":0.096,"uptime":99.1,"tps":46,"ttft":843,"reqs":984},{"name":"Novita","ctx":1048576,"in":0.48024,"out":0.96048,"cacheRead":0.003956,"uptime":97.2,"noSchema":1,"tps":26,"ttft":4503,"reqs":293},{"name":"StreamLake","ctx":1000000,"maxOut":128000,"in":0.522,"out":1.044,"cacheRead":0.00432,"uptime":94.6,"noSchema":1,"tps":17,"ttft":1644,"reqs":151}],"scores":{"arenaChat":{"v":1467.419793477693,"ci":3.8,"rank":44,"votes":60919,"as":"mimo-v2.5-pro","effort":""},"arenaWebdev":{"v":1474.9239949836478,"ci":5.9,"rank":50,"votes":15843,"as":"mimo-v2.5-pro","effort":""},"eqWriting":{"v":1490.5,"as":"XiaomiMiMo/MiMo-V2.5-Pro","version":"Creative Writing v3"},"eqChat":{"v":1208.3,"as":"xiaomi/mimo-v2.5-pro","version":"EQ-Bench 4","ciLow":1197.6,"ciHigh":1219,"boardAsOf":"2026-07-26T01:26:10+00:00"},"eqLongform":{"v":73.7,"as":"XiaomiMiMo/MiMo-V2.5-Pro","version":"Longform Writing v1.11"}},"perf":{"uptime":99.2,"cache":0.0028,"n":7,"tps":25,"ttft":3419,"tpsBest":46,"via":"Xiaomi","reqs":13422},"open":true,"rankable":true,"scored":5},{"id":"google/gemma-3-27b-it","name":"Google: Gemma 3 27B","org":"Google","slug":"gemma-3-27b-it","matchKey":"gemma327bit","blurb":"Gemma 3 introduces multimodality, supporting vision-language input and text outputs.","created":1741756359,"cutoff":"2024-08-31","hf":"google/gemma-3-27b-it","ctx":131072,"maxOut":117964,"in":["text","image"],"out":["text"],"price":{"in":0.08,"out":0.45},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.08,"out":0.16,"uptime":99.7,"tps":9,"ttft":1301,"reqs":10061},{"name":"Parasail","quant":"fp8","in":0.08,"out":0.45,"cacheRead":0.04,"uptime":99.6,"noTools":1,"tps":31,"ttft":758,"reqs":29235},{"name":"Nebius","quant":"fp8","ctx":110000,"maxOut":99000,"in":0.1,"out":0.3,"uptime":73.9,"noTools":1,"tps":28,"ttft":1020,"reqs":11448},{"name":"Novita","quant":"bf16","ctx":98304,"maxOut":16384,"in":0.119,"out":0.2,"uptime":86.6,"noTools":1,"noSchema":1,"tps":20,"ttft":1305,"reqs":4156}],"scores":{"arenaChat":{"v":1365.1873899533437,"ci":3.7,"rank":187,"votes":46422,"as":"gemma-3-27b-it","effort":""},"arenaVision":{"v":1158.7997230914252,"ci":8.2,"rank":99,"votes":16992,"as":"gemma-3-27b-it","effort":""},"bfcl":{"v":0.29469999999999996,"as":"Gemma-3-27b-it (Prompt)","effort":"","version":"V4","agent":"BFCL V4 · prompt-based tool calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1262.9,"as":"google/gemma-3-27b-it","version":"Creative Writing v3"},"eqLongform":{"v":41.9,"as":"google/gemma-3-27b-it","version":"Longform Writing v1.11"}},"perf":{"uptime":99.7,"n":4,"tps":31,"ttft":758,"via":"Parasail","reqs":29235},"open":true,"prose":{"slop":32.26903897173587,"trigram":0.7818818284097658,"notXButY":0.577146282667755,"repetition":10.171900247240446,"mattr":0.5686601691621295,"grade":9.885757791741515,"sentence":13.437539234149403,"words":335089,"samples":300,"as":"google/gemma-3-27b-it"},"rankable":true,"scored":5},{"id":"google/gemini-3.1-flash-lite-preview","name":"Google: Gemini 3.1 Flash Lite Preview","org":"Google","slug":"gemini-3.1-flash-lite-preview","matchKey":"gemini31flashlitepreview","blurb":"Gemini 3.1 Flash Lite Preview is Google's high-efficiency model optimized for high-volume use cases.","created":1772512673,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.25,"out":1.5,"imageIn":2.5e-7,"audioIn":0.5,"webSearch":0.014},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Google AI Studio","in":0.125,"out":0.75,"cacheRead":0.0125,"cacheWrite":0.041667,"uptime":100},{"name":"Google AI Studio","in":0.25,"out":1.5,"cacheRead":0.025,"cacheWrite":0.083333,"uptime":99.7},{"name":"Google AI Studio","in":0.45,"out":2.7,"cacheRead":0.045,"cacheWrite":0.15,"uptime":99.9,"tps":179,"ttft":425,"reqs":75178}],"scores":{"arenaChat":{"v":1432.3597744410254,"ci":3.7,"rank":98,"votes":60405,"as":"gemini-3.1-flash-lite-preview","effort":""},"arenaWebdev":{"v":1254.2676816579017,"ci":7,"rank":114,"votes":13634,"as":"gemini-3.1-flash-lite-preview","effort":""},"arenaVision":{"v":1234.9722797857262,"ci":6,"rank":57,"votes":25958,"as":"gemini-3.1-flash-lite-preview","effort":""},"arenaImageCode":{"v":1336.1618827284651,"ci":11,"rank":49,"votes":6095,"as":"gemini-3.1-flash-lite-preview","effort":""},"eqLongform":{"v":49,"as":"gemini-3.1-flash-lite-preview","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0125,"n":3,"tps":179,"ttft":425,"via":"Google AI Studio","reqs":75178},"open":false,"rankable":true,"scored":5},{"id":"anthropic/claude-sonnet-4","name":"Anthropic: Claude Sonnet 4","org":"Anthropic","slug":"claude-sonnet-4","matchKey":"claudesonnet4","blurb":"Claude Sonnet 4 significantly enhances the capabilities of its predecessor, Sonnet 3.7, excelling in both coding and reasoning tasks with improved precision and controllability.","created":1747930371,"cutoff":"2025-01-31","hf":null,"ctx":200000,"maxOut":64000,"in":["image","text","file"],"out":["text"],"price":{"in":3,"out":15,"webSearch":0.01},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"Amazon Bedrock","region":"eu","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":100},{"name":"Amazon Bedrock","in":3,"out":15,"cacheRead":0.3,"cacheWrite":3.75,"uptime":100,"tps":38,"ttft":878,"reqs":4578,"region":"eu"}],"scores":{"arenaChat":{"v":1400.9171097106896,"ci":4.5,"rank":148,"votes":33944,"as":"claude-sonnet-4-20250514-thinking-32k","effort":"default"},"arenaVision":{"v":1206.0855119112775,"ci":16.9,"rank":75,"votes":1145,"as":"claude-sonnet-4-20250514-thinking-32k","effort":"default"},"swebenchAgent":{"v":0.768,"as":"Claude 4 Sonnet","agent":"EPAM AI/Run Developer Agent","date":"2025-08-04","checked":false},"eqWriting":{"v":1480.3,"as":"claude-sonnet-4","version":"Creative Writing v3"},"eqLongform":{"v":58.4,"as":"claude-sonnet-4","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.3,"regions":["eu"],"n":2,"tps":38,"ttft":878,"via":"Amazon Bedrock","reqs":4578},"open":false,"rankable":true,"scored":5},{"id":"deepseek/deepseek-r1","name":"DeepSeek: R1","org":"DeepSeek","slug":"deepseek-r1","matchKey":"deepseekr1","blurb":"DeepSeek R1 is here: Performance on par with [OpenAI o1](/openai/o1), but open-sourced and with fully open reasoning tokens.","created":1737381095,"cutoff":"2024-07-31","hf":"deepseek-ai/DeepSeek-R1","ctx":64000,"maxOut":16000,"in":["text"],"out":["text"],"price":{"in":0.7,"out":2.5},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"fp8","in":0.7,"out":2.5,"uptime":100,"tps":23,"ttft":1123,"reqs":848}],"scores":{"arenaChat":{"v":1398.035776349719,"ci":4.9,"rank":151,"votes":18524,"as":"deepseek-r1","effort":""},"aime":{"v":0.6638888888888889,"as":"DeepSeek-R1 (May 2025)","effort":"","date":"2025-05-29","stderr":0.04911177955575991},"gpqa":{"v":0.7632575757575758,"as":"DeepSeek-R1 (May 2025)","effort":"","date":"2025-05-29","stderr":0.024023818843083005},"eqWriting":{"v":1500,"as":"deepseek-ai/DeepSeek-R1","version":"Creative Writing v3"},"eqLongform":{"v":59.5,"as":"deepseek-ai/DeepSeek-R1","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":1,"tps":23,"ttft":1123,"via":"Novita","reqs":848},"open":true,"prose":{"slop":40.16025438049749,"trigram":0.6999463374474624,"notXButY":0.13681829789146704,"repetition":5.125672264295906,"mattr":0.6176627048249442,"grade":9.359502761775467,"sentence":12.531071153143646,"words":300023,"samples":299,"as":"deepseek/deepseek-r1"},"rankable":true,"scored":5},{"id":"openai/o3-mini","name":"OpenAI: o3 Mini","org":"OpenAI","slug":"o3-mini","matchKey":"o3mini","blurb":"OpenAI o3-mini is a cost-efficient language model optimized for STEM reasoning tasks, particularly excelling in science, mathematics, and coding.","created":1738351721,"cutoff":"2023-10-31","hf":null,"ctx":200000,"maxOut":100000,"in":["text","file"],"out":["text"],"price":{"in":1.1,"out":4.4,"webSearch":0.01,"alt":{"batch":{"in":0.55,"out":2.2}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":1.1,"out":4.4,"cacheRead":0.55,"uptime":100,"tps":109,"ttft":1667,"reqs":20}],"scores":{"frontierMath":{"v":0.18596491228070175,"as":"o3-mini (high)","effort":"high","date":"2026-06-11","stderr":0.023087552563757593},"aime":{"v":0.7694444444444445,"as":"o3-mini (high)","effort":"high","date":"2025-02-27","stderr":0.05239762076828802},"gpqa":{"v":0.7702020202020202,"as":"o3-mini (high)","effort":"high","date":"2025-02-13","stderr":0.025855332698878485},"simpleqa":{"v":0.153,"as":"o3-mini (high)","effort":"high","date":"2026-08-31","stderr":0.011389500459665504},"swebenchAgent":{"v":0.424,"as":"O3 Mini","agent":"Agentless Lite","date":"2025-02-14","checked":true}},"perf":{"uptime":100,"cache":0.55,"n":1,"tps":109,"ttft":1667,"via":"OpenAI","reqs":20},"open":false,"rankable":true,"scored":5},{"id":"qwen/qwen3-max","name":"Qwen: Qwen3 Max","org":"Qwen","slug":"qwen3-max","matchKey":"qwen3max","blurb":"Qwen3-Max is an updated release built on the Qwen3 series, offering major improvements in reasoning, instruction following, multilingual support, and long-tail knowledge coverage compared to the January 2025 version.","created":1758662808,"cutoff":"2025-06-30","hf":null,"ctx":262144,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.78,"out":3.9},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.78,"out":3.9,"cacheRead":0.156,"cacheWrite":0.975,"uptime":100,"tps":38,"ttft":693,"reqs":2030}],"scores":{"arenaChat":{"v":1434.683579379916,"ci":4.5,"rank":94,"votes":27194,"as":"qwen3-max-preview","effort":"max"},"frontierMath":{"v":0.18947368421052632,"as":"Qwen3-Max-Instruct","effort":"max","date":"2026-08-30","stderr":0.023254062737737245},"aime":{"v":0.7333333333333333,"as":"Qwen3-Max-Instruct","effort":"max","date":"2025-10-06","stderr":0.059139451700829515},"gpqa":{"v":0.726010101010101,"as":"Qwen3-Max-Instruct","effort":"max","date":"2025-10-06","stderr":0.02734528103171235},"simpleqa":{"v":0.4874874874874875,"as":"Qwen3-Max-Instruct","effort":"max","date":"2026-08-27","stderr":0.01582226675546796}},"perf":{"uptime":100,"cache":0.156,"n":1,"tps":38,"ttft":693,"via":"Alibaba","reqs":2030},"open":false,"rankable":true,"scored":5},{"id":"z-ai/glm-5v-turbo","name":"Z.ai: GLM 5V Turbo","org":"Z.ai","slug":"glm-5v-turbo","matchKey":"glm5vturbo","blurb":"GLM-5V-Turbo is Z.ai’s first native multimodal agent foundation model, built for vision-based coding and agent-driven tasks.","created":1775061458,"cutoff":null,"hf":null,"ctx":202752,"maxOut":131072,"in":["image","text","video"],"out":["text"],"price":{"in":1.2,"out":4},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Z.AI","quant":"fp8","in":1.2,"out":4,"cacheRead":0.24,"uptime":100,"tps":16,"ttft":8500,"reqs":162}],"scores":{"arenaChat":{"v":1432.6771549776568,"ci":6.7,"rank":97,"votes":9361,"as":"glm-5v-turbo","effort":""},"arenaWebdev":{"v":1400.6851493406753,"ci":13.9,"rank":75,"votes":2185,"as":"glm-5v-turbo","effort":""},"arenaVision":{"v":1228.3305918528506,"ci":5.7,"rank":61,"votes":37424,"as":"glm-5v-turbo","effort":""},"arenaDocument":{"v":1415.5933758703818,"ci":8.2,"rank":39,"votes":6995,"as":"glm-5v-turbo","effort":""},"arenaImageCode":{"v":1423.0044261924972,"ci":16.6,"rank":42,"votes":1444,"as":"glm-5v-turbo","effort":""}},"perf":{"uptime":100,"cache":0.24,"n":1,"tps":16,"ttft":8500,"via":"Z.AI","reqs":162},"open":false,"rankable":true,"scored":5},{"id":"deepseek/deepseek-v4-pro-0813","name":"DeepSeek: DeepSeek V4 Pro 0813","org":"DeepSeek","slug":"deepseek-v4-pro-0813","matchKey":"deepseekv4pro0813","blurb":"DeepSeek V4 Pro 0813 is a large-scale mixture-of-experts model from DeepSeek.","created":1786549364,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V4-Pro-0813","ctx":1048576,"maxOut":384000,"in":["text"],"out":["text"],"price":{"in":1.32,"out":3.96,"alt":{"batch":{"in":0.66,"out":1.98}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"StreamLake","ctx":1024000,"in":0.558624,"out":1.675872,"cacheRead":0.018621,"uptime":99.8,"noSchema":1,"tps":44,"ttft":2962,"reqs":12746},{"name":"Wafer","maxOut":943718,"in":0.57,"out":2.9,"cacheRead":0.132,"uptime":98.6,"tps":88,"ttft":871,"reqs":8429},{"name":"Ionstream","maxOut":943718,"in":0.7,"out":2.88,"cacheRead":0.088,"uptime":99.3,"tps":95,"ttft":730,"reqs":941},{"name":"Sail Research","region":"us","quant":"fp4","in":0.7,"out":2.96,"cacheRead":0.033,"uptime":99.3},{"name":"Sail Research","quant":"fp4","in":0.7,"out":2.96,"cacheRead":0.033,"uptime":99.5,"tps":13,"ttft":1851,"reqs":456,"region":"us"},{"name":"Novita","quant":"fp8","maxOut":393216,"in":0.99,"out":2.97,"cacheRead":0.033,"uptime":100,"noSchema":1,"tps":55,"ttft":1247,"reqs":464},{"name":"GMICloud","quant":"fp8","ctx":1048575,"maxOut":943717,"in":1.056,"out":3.168,"cacheRead":0.0352,"uptime":99.9,"noSchema":1,"tps":40,"ttft":4164,"reqs":156},{"name":"NextBit","quant":"fp8","maxOut":943718,"in":1.122,"out":3.366,"cacheRead":0.037,"uptime":98.3,"tps":39,"ttft":2943,"reqs":126},{"name":"Alibaba","ctx":1000000,"maxOut":393216,"in":1.122,"out":3.366,"cacheRead":0.1122,"uptime":100,"tps":67,"ttft":1190,"reqs":607},{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":1.3,"out":2.6,"cacheRead":0.1,"uptime":99.2,"tps":67,"ttft":931,"reqs":779},{"name":"CoreWeave","quant":"fp8","maxOut":943718,"in":1.31,"out":3.96,"cacheRead":0.044,"uptime":99.6,"tps":81,"ttft":824,"reqs":853},{"name":"AtlasCloud","quant":"fp8","maxOut":393216,"in":1.32,"out":3.96,"cacheRead":0.132,"uptime":99.5,"tps":53,"ttft":1422,"reqs":88},{"name":"Baidu","quant":"fp8","maxOut":393216,"in":1.32,"out":3.96,"cacheRead":0.042,"uptime":100,"tps":48,"ttft":1082,"reqs":768},{"name":"BaseTen","quant":"fp4","maxOut":262144,"in":1.32,"out":3.96,"cacheRead":0.132,"uptime":97.9},{"name":"Parasail","quant":"fp8","maxOut":943718,"in":1.32,"out":3.96,"cacheRead":0.044,"uptime":99.7,"tps":67,"ttft":809,"reqs":120},{"name":"Together","maxOut":943718,"in":1.32,"out":3.96,"cacheRead":0.13,"uptime":99.5,"tps":41,"ttft":924,"reqs":237},{"name":"DigitalOcean","maxOut":943718,"in":1.32,"out":3.96,"cacheRead":0.044,"uptime":99.7,"noSchema":1,"tps":31,"ttft":863,"reqs":78},{"name":"SiliconFlow","quant":"fp8","maxOut":393216,"in":1.32,"out":3.96,"cacheRead":0.044,"uptime":98.5,"noSchema":1,"tps":45,"ttft":2037,"reqs":80},{"name":"BaseTen","quant":"fp4","maxOut":262144,"in":1.32,"out":3.96,"cacheRead":0.132,"uptime":97.6,"tps":88,"ttft":575,"reqs":773,"region":"us"},{"name":"Cloudflare","maxOut":943718,"in":1.32,"out":3.96,"cacheRead":0.044,"uptime":97.3,"tps":32,"ttft":1739,"reqs":51},{"name":"Fireworks","maxOut":943718,"in":1.32,"out":3.96,"cacheRead":0.044,"uptime":97.8,"tps":40,"ttft":949,"reqs":76},{"name":"DeepSeek","in":1.32,"out":3.96,"cacheRead":0.044,"uptime":100,"noSchema":1,"tps":35,"ttft":1213,"reqs":11593},{"name":"Phala","maxOut":393216,"in":1.45,"out":4.36,"cacheRead":0.15,"uptime":99.7,"tps":48,"ttft":1844,"reqs":83},{"name":"Venice","ctx":1000000,"maxOut":32768,"in":1.65,"out":4.95,"cacheRead":0.165,"uptime":98.6,"tps":74,"ttft":631,"reqs":386}],"scores":{"frontierMath":{"v":0.6456140350877193,"as":"DeepSeek V4 Pro 0813 (max)","effort":"max","date":"2026-08-19","stderr":0.028383475205435625},"aime":{"v":0.9861111111111112,"as":"DeepSeek V4 Pro 0813 (max)","effort":"max","date":"2026-08-18","stderr":0.00990980650017605},"gpqa":{"v":0.9166666666666666,"as":"DeepSeek V4 Pro 0813 (max)","effort":"max","date":"2026-08-18","stderr":0.015242573629934338},"simpleqa":{"v":0.5290581162324649,"as":"DeepSeek V4 Pro 0813 (max)","effort":"max","date":"2026-08-27","stderr":0.015808394665706597}},"perf":{"uptime":100,"cache":0.018621,"regions":["us"],"n":24,"tps":44,"ttft":2962,"tpsBest":95,"via":"StreamLake","reqs":12746},"open":true,"rankable":true,"scored":4},{"id":"deepseek/deepseek-v4-flash","name":"DeepSeek: DeepSeek V4 Flash 0423","org":"DeepSeek","slug":"deepseek-v4-flash","matchKey":"deepseekv4flash","blurb":"DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window.","created":1777000666,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V4-Flash","ctx":1048576,"maxOut":384000,"in":["text"],"out":["text"],"price":{"in":0.088606,"out":0.177212},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high"]},"moderated":false,"providers":[{"name":"OpenInference","quant":"fp8","maxOut":943718,"in":0.04,"out":0.3,"cacheRead":0.014,"uptime":99.5,"tps":19,"ttft":1282,"reqs":150925},{"name":"StreamLake","quant":"fp8","ctx":1024000,"in":0.088606,"out":0.177212,"cacheRead":0.017721,"uptime":98,"tps":45,"ttft":1366,"reqs":125093},{"name":"DeepInfra","quant":"fp8","maxOut":65536,"in":0.09,"out":0.18,"cacheRead":0.018,"uptime":99.6,"tps":16,"ttft":4675,"reqs":4705},{"name":"GMICloud","quant":"fp8","ctx":1048575,"maxOut":943717,"in":0.091,"out":0.182,"cacheRead":0.0182,"uptime":99.7,"noSchema":1,"tps":38,"ttft":2196,"reqs":58526},{"name":"Venice","ctx":1000000,"maxOut":32768,"in":0.0966,"out":0.1925,"cacheRead":0.0196,"uptime":98.1,"tps":18,"ttft":4757,"reqs":34068},{"name":"DigitalOcean","in":0.098,"out":0.196,"cacheRead":0.0196,"uptime":99.9,"tps":16,"ttft":940,"reqs":30555},{"name":"SiliconFlow","quant":"fp8","maxOut":393216,"in":0.13,"out":0.28,"cacheRead":0.028,"uptime":98.9,"noSchema":1,"tps":23,"ttft":1285,"reqs":9633},{"name":"Alibaba","quant":"fp8","ctx":1000000,"maxOut":393216,"in":0.134,"out":0.268,"cacheRead":0.0268,"uptime":99.2,"tps":83,"ttft":896,"reqs":70091},{"name":"Baidu","quant":"fp8","maxOut":131072,"in":0.14,"out":0.28,"cacheRead":0.028,"uptime":85.1,"tps":91,"ttft":809,"reqs":2991},{"name":"Novita","quant":"fp8","maxOut":393216,"in":0.14,"out":0.28,"cacheRead":0.028,"uptime":100,"noSchema":1,"tps":55,"ttft":1210,"reqs":136520},{"name":"AtlasCloud","quant":"fp4","maxOut":393216,"in":0.14,"out":0.28,"cacheRead":0.028,"uptime":99.9,"tps":31,"ttft":1039,"reqs":87715},{"name":"Parasail","quant":"fp8","maxOut":943718,"in":0.14,"out":0.28,"cacheRead":0.07,"uptime":99.6,"tps":56,"ttft":799,"reqs":37000},{"name":"NextBit","quant":"fp8","maxOut":943718,"in":0.15,"out":0.3,"cacheRead":0.035,"uptime":98.5,"tps":28,"ttft":1910,"reqs":46432},{"name":"Mancer 2","quant":"fp8","maxOut":943718,"in":0.19,"out":0.5,"uptime":97.9,"tps":44,"ttft":910,"reqs":3394},{"name":"Azure","region":"us","in":0.21,"out":0.56,"cacheRead":0.031,"uptime":95.4,"noSchema":1,"tps":52,"ttft":1439,"reqs":8945}],"scores":{"arenaChat":{"v":1437.93628921571,"ci":4.2,"rank":87,"votes":48597,"as":"deepseek-v4-flash-high-preview","effort":"high"},"arenaWebdev":{"v":1580.2013421567158,"ci":9.5,"rank":22,"votes":4723,"as":"deepseek-v4-flash-high","effort":"high"},"eqWriting":{"v":1555.7,"as":"deepseek-ai/DeepSeek-V4-Flash","version":"Creative Writing v3"},"eqLongform":{"v":66,"as":"deepseek-ai/DeepSeek-V4-Flash","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.014,"regions":["us"],"n":15,"tps":19,"ttft":1282,"tpsBest":91,"via":"OpenInference","reqs":150925},"open":true,"rankable":true,"scored":4},{"id":"qwen/qwen3.5-27b","name":"Qwen: Qwen3.5-27B","org":"Qwen","slug":"qwen3.5-27b","matchKey":"qwen3527b","blurb":"The Qwen3.5 27B native vision-language Dense model incorporates a linear attention mechanism, delivering fast response times while balancing inference speed and performance.","created":1772053810,"cutoff":null,"hf":"Qwen/Qwen3.5-27B","ctx":262144,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.195,"out":1.56},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.195,"out":1.56,"uptime":99,"tps":19,"ttft":989,"reqs":3984},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.25,"out":2,"uptime":94.6,"tps":13,"ttft":919,"reqs":1863},{"name":"DeepInfra","quant":"fp8","maxOut":81920,"in":0.26,"out":2.6,"uptime":99.2,"tps":36,"ttft":483,"reqs":1839},{"name":"AtlasCloud","quant":"fp8","in":0.27,"out":2.16,"cacheRead":0.27,"uptime":99.5,"tps":19,"ttft":1334,"reqs":2138},{"name":"Phala","in":0.3,"out":2.4,"cacheRead":0.15,"uptime":88.8,"noTools":1,"tps":12,"ttft":1536,"reqs":1572},{"name":"Novita","quant":"bf16","in":0.3,"out":2.4,"uptime":99.8,"noSchema":1,"tps":16,"ttft":1228,"reqs":1354}],"scores":{"arenaChat":{"v":1407.880961593894,"ci":4.4,"rank":140,"votes":27227,"as":"qwen3.5-27b","effort":""},"arenaWebdev":{"v":1357.376497113762,"ci":8.2,"rank":95,"votes":7427,"as":"qwen3.5-27b","effort":""},"arenaVision":{"v":1217.675736837781,"ci":6,"rank":66,"votes":25350,"as":"qwen3.5-27b","effort":""},"eqLongform":{"v":59,"as":"Qwen/Qwen3.5-27B","version":"Longform Writing v1.11"}},"perf":{"uptime":99.8,"n":6,"tps":19,"ttft":989,"tpsBest":36,"via":"Alibaba","reqs":3984},"open":true,"rankable":true,"scored":4},{"id":"qwen/qwen3.6-27b","name":"Qwen: Qwen3.6 27B","org":"Qwen","slug":"qwen3.6-27b","matchKey":"qwen3627b","blurb":"Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.","created":1777255064,"cutoff":null,"hf":"Qwen/Qwen3.6-27B","ctx":262144,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.3,"out":2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Chutes","quant":"fp8","in":0.3,"out":2,"cacheRead":0.03,"uptime":94.2,"tps":20,"ttft":2727,"reqs":897},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.3,"out":3.2,"uptime":79.9,"tps":8,"ttft":5284,"reqs":25},{"name":"Phala","maxOut":262140,"in":0.32,"out":2.7,"cacheRead":0.15,"uptime":97.4,"tps":27,"ttft":1225,"reqs":2368},{"name":"DeepInfra","quant":"fp8","maxOut":81920,"in":0.32,"out":3.2,"uptime":99.6,"tps":38,"ttft":802,"reqs":566},{"name":"Venice","quant":"fp8","ctx":256000,"in":0.325,"out":3.25,"uptime":98.6,"tps":36,"ttft":602,"reqs":407},{"name":"Alibaba","in":0.45,"out":2.7,"uptime":99.8,"tps":44,"ttft":774,"reqs":198}],"scores":{"frontierMath":{"v":0.3508771929824561,"as":"Qwen3.6 27B","effort":"","date":"2026-08-28","stderr":0.028319255656547093},"aime":{"v":0.9111111111111111,"as":"Qwen3.6 27B","effort":"","date":"2026-08-07","stderr":0.04290254662948546},"gpqa":{"v":0.8585858585858586,"as":"Qwen3.6 27B","effort":"","date":"2026-08-07","stderr":0.024825909793343353},"eqChat":{"v":1025.5,"as":"qwen/qwen3.6-27b","version":"EQ-Bench 4","ciLow":1005.8,"ciHigh":1046.3,"boardAsOf":"2026-07-26T01:26:10+00:00"}},"perf":{"uptime":99.8,"cache":0.03,"n":6,"tps":27,"ttft":1225,"tpsBest":44,"via":"Phala","reqs":2368},"open":true,"rankable":true,"scored":4},{"id":"qwen/qwen3.5-122b-a10b","name":"Qwen: Qwen3.5-122B-A10B","org":"Qwen","slug":"qwen3.5-122b-a10b","matchKey":"qwen35122ba10b","blurb":"The Qwen3.5 122B-A10B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency.","created":1772053789,"cutoff":null,"hf":"Qwen/Qwen3.5-122B-A10B","ctx":262144,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.26,"out":2.08},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.26,"out":2.08,"uptime":90.1,"noTools":1,"tps":30,"ttft":783,"reqs":2763},{"name":"Alibaba","in":0.26,"out":2.08,"uptime":96.5,"tps":98,"ttft":1520,"reqs":43},{"name":"DeepInfra","quant":"fp4","maxOut":81920,"in":0.29,"out":2.4,"uptime":99.9,"noSchema":1,"tps":93,"ttft":452,"reqs":482},{"name":"AtlasCloud","quant":"fp8","in":0.3,"out":2.4,"cacheRead":0.3,"uptime":99,"tps":64,"ttft":1695,"reqs":110},{"name":"Novita","quant":"bf16","in":0.4,"out":3.2,"uptime":99.7,"noSchema":1,"tps":45,"ttft":1500,"reqs":115}],"scores":{"arenaChat":{"v":1416.777978713208,"ci":4.4,"rank":124,"votes":28359,"as":"qwen3.5-122b-a10b","effort":""},"arenaWebdev":{"v":1357.9329031686448,"ci":7.8,"rank":94,"votes":7831,"as":"qwen3.5-122b-a10b","effort":""},"arenaVision":{"v":1226.6249795551641,"ci":7,"rank":62,"votes":12465,"as":"qwen3.5-122b-a10b","effort":""},"eqLongform":{"v":54.3,"as":"Qwen/Qwen3.5-122B-A10B","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"n":5,"tps":30,"ttft":783,"tpsBest":98,"via":"SiliconFlow","reqs":2763},"open":true,"rankable":true,"scored":4},{"id":"deepseek/deepseek-v3.2-exp","name":"DeepSeek: DeepSeek V3.2 Exp","org":"DeepSeek","slug":"deepseek-v3.2-exp","matchKey":"deepseekv32exp","blurb":"DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures.","created":1759150481,"cutoff":"2025-07-31","hf":"deepseek-ai/DeepSeek-V3.2-Exp","ctx":163840,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.27,"out":0.41},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"SiliconFlow","quant":"fp8","maxOut":147456,"in":0.27,"out":0.41,"uptime":96.8,"tps":12,"ttft":2257,"reqs":866},{"name":"AtlasCloud","quant":"fp8","maxOut":147456,"in":0.27,"out":0.41,"cacheRead":0.27,"uptime":99.3,"tps":46,"ttft":1764,"reqs":385},{"name":"Novita","quant":"fp8","in":0.27,"out":0.41,"uptime":99.4,"tps":34,"ttft":2095,"reqs":1513}],"scores":{"arenaChat":{"v":1425.279766308678,"ci":6.6,"rank":110,"votes":8905,"as":"deepseek-v3.2-exp-thinking","effort":"default"},"arenaWebdev":{"v":1271.7475429543197,"ci":13.6,"rank":111,"votes":3865,"as":"deepseek-v3.2-exp","effort":""},"bfcl":{"v":0.5412,"as":"DeepSeek-V3.2-Exp (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqLongform":{"v":60.2,"as":"deepseek-ai/DeepSeek-V3.2-Exp","version":"Longform Writing v1.11"}},"perf":{"uptime":99.4,"n":3,"tps":34,"ttft":2095,"tpsBest":46,"via":"Novita","reqs":1513},"open":true,"prose":{"slop":23.950797771784337,"trigram":0.5217654324099513,"notXButY":0.5690958453957651,"repetition":4.838402783440702,"mattr":0.5615172109197558,"grade":9.459844321051854,"sentence":15.209866252415413,"words":396730,"samples":300,"as":"deepseek/deepseek-v3.2-exp"},"rankable":true,"scored":4},{"id":"meta-llama/llama-4-scout","name":"Meta: Llama 4 Scout","org":"Meta","slug":"llama-4-scout","matchKey":"llama4scout","blurb":"Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B.","created":1743881519,"cutoff":"2024-08-31","hf":"meta-llama/Llama-4-Scout-17B-16E-Instruct","ctx":1310720,"maxOut":16384,"in":["text","image"],"out":["text"],"price":{"in":0.1,"out":0.3},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","ctx":327680,"in":0.1,"out":0.3,"uptime":99.9,"noTools":1,"tps":16,"ttft":749,"reqs":13377},{"name":"Novita","quant":"bf16","ctx":131072,"maxOut":117964,"in":0.18,"out":0.59,"uptime":99.9,"noTools":1,"noSchema":1,"tps":24,"ttft":753,"reqs":2614},{"name":"Google","region":"us","maxOut":8192,"in":0.25,"out":0.7}],"scores":{"swebenchAgent":{"v":0.0906,"as":"Llama 4 Scout Instruct","agent":"mini-SWE-agent","date":"2025-07-20","checked":true},"bfcl":{"v":0.2813,"as":"Llama-4-Scout-17B-16E-Instruct (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":781.2,"as":"meta-llama/Llama-4-Scout-17B-16E-Instruct","version":"Creative Writing v3"},"eqLongform":{"v":31.1,"as":"meta-llama/Llama-4-Scout-17B-16E-Instruct","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"regions":["us"],"n":3,"tps":16,"ttft":749,"tpsBest":24,"via":"DeepInfra","reqs":13377},"open":true,"rankable":true,"scored":4},{"id":"mistralai/mistral-medium-3-5","name":"Mistral: Mistral Medium 3.5","org":"Mistral","slug":"mistral-medium-3-5","matchKey":"mistralmedium35","blurb":"Mistral Medium 3.5 is a dense 128B instruction-following model from Mistral AI.","created":1777570439,"cutoff":null,"hf":null,"ctx":262144,"maxOut":209715,"in":["text","image","file"],"out":["text"],"price":{"in":1.5,"out":7.5,"alt":{"batch":{"in":0.75,"out":3.75}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","none"]},"moderated":false,"providers":[{"name":"Mistral","in":1.5,"out":7.5,"uptime":99.8},{"name":"Mistral","in":1.5,"out":7.5,"uptime":99.8,"tps":64,"ttft":252,"reqs":263}],"scores":{"arenaChat":{"v":1426.3465450999593,"ci":6.6,"rank":106,"votes":10996,"as":"mistral-medium-3.5","effort":"medium"},"arenaWebdev":{"v":1264.9163780310282,"ci":15.3,"rank":112,"votes":2154,"as":"mistral-medium-3.5","effort":"medium"},"arenaVision":{"v":1199.0449789106383,"ci":9.9,"rank":80,"votes":5090,"as":"mistral-medium-3.5","effort":"medium"},"eqChat":{"v":993,"as":"mistralai/mistral-medium-3-5","version":"EQ-Bench 4","ciLow":974.4,"ciHigh":1011.5,"boardAsOf":"2026-07-26T01:26:10+00:00"}},"perf":{"uptime":99.8,"n":2,"tps":64,"ttft":252,"via":"Mistral","reqs":263},"open":false,"rankable":true,"scored":4},{"id":"anthropic/claude-opus-4.1","name":"Anthropic: Claude Opus 4.1","org":"Anthropic","slug":"claude-opus-4.1","matchKey":"claudeopus41","blurb":"Claude Opus 4.1 is an updated version of Anthropic’s flagship model, offering improved performance in coding, reasoning, and agentic tasks.","created":1754411591,"cutoff":"2025-01-31","hf":null,"ctx":200000,"maxOut":32000,"in":["image","text","file"],"out":["text"],"price":{"in":15,"out":75,"webSearch":0.01,"alt":{"batch":{"in":7.5,"out":37.5}}},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"Amazon Bedrock","in":15,"out":75,"cacheRead":1.5,"cacheWrite":18.75,"uptime":100,"tps":6,"ttft":4164,"reqs":36}],"scores":{"arenaChat":{"v":1449.8517999878472,"ci":3.5,"rank":70,"votes":48811,"as":"claude-opus-4-1-20250805-thinking-16k","effort":"default"},"arenaWebdev":{"v":1388.969799226031,"ci":10.9,"rank":81,"votes":6769,"as":"claude-opus-4-1-20250805","effort":""},"arenaSearch":{"v":1148.2792594379139,"ci":4.9,"rank":25,"votes":76933,"as":"claude-opus-4-1-search","effort":""},"eqLongform":{"v":70.5,"as":"claude-opus-4.1","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":1.5,"n":1,"tps":6,"ttft":4164,"via":"Amazon Bedrock","reqs":36},"open":false,"rankable":true,"scored":4},{"id":"moonshotai/kimi-k2","name":"MoonshotAI: Kimi K2 0711","org":"MoonshotAI","slug":"kimi-k2","matchKey":"kimik2","blurb":"Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass.","created":1752263252,"cutoff":"2024-12-31","hf":"moonshotai/Kimi-K2-Instruct","ctx":131072,"maxOut":98304,"in":["text"],"out":["text"],"price":{"in":0.57,"out":2.3},"caps":{"tools":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Novita","quant":"fp8","in":0.57,"out":2.3,"uptime":100,"tps":26,"ttft":1240,"reqs":493}],"scores":{"swebenchAgent":{"v":0.7120000000000001,"as":"Kimi K2","agent":"Lingxi v1.5","date":"2025-10-14","checked":false},"bfcl":{"v":0.5906,"as":"Moonshotai-Kimi-K2-Instruct (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1662.7,"as":"moonshotai/Kimi-K2-Instruct","version":"Creative Writing v3"},"eqLongform":{"v":68.1,"as":"moonshotai/Kimi-K2-Instruct","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":1,"tps":26,"ttft":1240,"via":"Novita","reqs":493},"open":true,"prose":{"slop":12.946588436217393,"trigram":0.135434169811494,"notXButY":0.1961294156977824,"repetition":4.136976389347679,"mattr":0.6221713345452644,"grade":9.384473195706615,"sentence":16.529168529280287,"words":435636,"samples":300,"as":"moonshotai/kimi-k2"},"rankable":true,"scored":4},{"id":"qwen/qwen3.6-flash","name":"Qwen: Qwen3.6 Flash","org":"Qwen","slug":"qwen3.6-flash","matchKey":"qwen36flash","blurb":"Qwen3.6 Flash is a fast, efficient language model from Alibaba's Qwen 3.6 series.","created":1777261362,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.1875,"out":1.125},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.1875,"out":1.125,"cacheWrite":0.234375,"uptime":100,"tps":76,"ttft":364,"reqs":3665}],"scores":{"frontierMath":{"v":0.22456140350877193,"as":"Qwen 3.6 Flash (2026-04-16)","effort":"","date":"2026-08-29","stderr":0.024761793289496806},"aime":{"v":0.8444444444444444,"as":"Qwen 3.6 Flash (2026-04-16)","effort":"","date":"2026-08-07","stderr":0.05463890236888291},"gpqa":{"v":0.8333333333333334,"as":"Qwen 3.6 Flash (2026-04-16)","effort":"","date":"2026-08-07","stderr":0.026552207828215258},"simpleqa":{"v":0.15915915915915915,"as":"Qwen 3.6 Flash (2026-04-16)","effort":"","date":"2026-08-27","stderr":0.011579970321213651}},"perf":{"uptime":100,"n":1,"tps":76,"ttft":364,"via":"Alibaba","reqs":3665},"open":false,"rankable":true,"scored":4},{"id":"z-ai/glm-4.5","name":"Z.ai: GLM 4.5","org":"Z.ai","slug":"glm-4.5","matchKey":"glm45","blurb":"GLM-4.5 is our latest flagship foundation model, purpose-built for agent-based applications.","created":1753471347,"cutoff":"2024-12-31","hf":"zai-org/GLM-4.5","ctx":131072,"maxOut":98304,"in":["text"],"out":["text"],"price":{"in":0.6,"out":2.2},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Z.AI","quant":"fp8","in":0.6,"out":2.2,"cacheRead":0.11,"uptime":100,"tps":19,"ttft":9150,"reqs":143}],"scores":{"arenaChat":{"v":1411.3028675112555,"ci":4.9,"rank":136,"votes":23712,"as":"glm-4.5","effort":""},"swebenchAgent":{"v":0.642,"as":"GLM-4.5","agent":"Undisclosed","date":"2025-07-28","checked":false},"eqWriting":{"v":1340.5,"as":"zai-org/GLM-4.5","version":"Creative Writing v3"},"eqLongform":{"v":55.5,"as":"zai-org/GLM-4.5","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.11,"n":1,"tps":19,"ttft":9150,"via":"Z.AI","reqs":143},"open":true,"prose":{"slop":28.446055294314608,"trigram":0.7900866158152431,"notXButY":0.42388987350114976,"repetition":5.365902492742016,"mattr":0.5722030120303201,"grade":10.47022636988056,"sentence":15.106828820158363,"words":340469,"samples":299,"as":"z-ai/glm-4.5"},"rankable":true,"scored":4},{"id":"openai/gpt-oss-120b","name":"OpenAI: gpt-oss-120b","org":"OpenAI","slug":"gpt-oss-120b","matchKey":"gptoss120b","blurb":"gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases.","created":1754414231,"cutoff":"2024-06-30","hf":"openai/gpt-oss-120b","ctx":131072,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.15,"out":0.6},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"AkashML","quant":"bf16","maxOut":117964,"in":0.03,"out":0.17,"cacheRead":0.03,"uptime":99.9,"tps":31,"ttft":776,"reqs":41394},{"name":"CoreWeave","quant":"fp4","maxOut":117964,"in":0.03,"out":0.17,"cacheRead":0.03,"uptime":99.9,"tps":49,"ttft":387,"reqs":52110},{"name":"DekaLLM","quant":"bf16","maxOut":117964,"in":0.03,"out":0.18,"uptime":99.7,"tps":31,"ttft":648,"reqs":22750},{"name":"DeepInfra","quant":"bf16","maxOut":117964,"in":0.037,"out":0.17,"uptime":99.7},{"name":"Crusoe","quant":"bf16","maxOut":117964,"in":0.05,"out":0.25,"cacheRead":0.05,"uptime":99.6,"tps":221,"ttft":177,"reqs":13872},{"name":"Novita","quant":"fp4","maxOut":32768,"in":0.05,"out":0.25,"uptime":92.5,"noSchema":1,"tps":83,"ttft":893,"reqs":61},{"name":"Mancer 2","quant":"fp8","maxOut":117964,"in":0.05,"out":0.3,"uptime":98.6,"tps":61,"ttft":529,"reqs":7468},{"name":"DigitalOcean","ctx":128000,"maxOut":4096,"in":0.06,"out":0.42,"cacheRead":0.012,"uptime":100,"noTools":1,"noSchema":1,"tps":29,"ttft":533,"reqs":17739},{"name":"Google","region":"global","maxOut":117964,"in":0.09,"out":0.36,"uptime":84.7,"noTools":1,"tps":140,"ttft":1740,"reqs":101505},{"name":"BaseTen","quant":"fp4","ctx":128072,"maxOut":115264,"in":0.1,"out":0.5,"cacheRead":0.1,"uptime":99.9},{"name":"BaseTen","quant":"fp4","ctx":128072,"maxOut":115264,"in":0.1,"out":0.5,"cacheRead":0.1,"uptime":100,"tps":190,"ttft":256,"reqs":4723,"region":"us"},{"name":"Parasail","quant":"fp4","maxOut":117964,"in":0.1,"out":0.75,"cacheRead":0.055,"uptime":100,"tps":100,"ttft":456,"reqs":3174},{"name":"SambaNova","maxOut":117964,"in":0.14,"out":0.95,"uptime":97.7,"noSchema":1,"tps":231,"ttft":753,"reqs":1152},{"name":"Amazon Bedrock","region":"eu","maxOut":117964,"in":0.15,"out":0.6,"uptime":100,"noTools":1,"noSchema":1},{"name":"Nebius","quant":"fp4","maxOut":117964,"in":0.15,"out":0.6,"uptime":99.5,"tps":268,"ttft":248,"reqs":3162},{"name":"Amazon Bedrock","maxOut":117964,"in":0.15,"out":0.6,"uptime":99.5,"noTools":1,"noSchema":1,"tps":90,"ttft":634,"reqs":13720},{"name":"DeepInfra","quant":"bf16","maxOut":16384,"in":0.15,"out":0.6,"uptime":100},{"name":"SiliconFlow","quant":"fp8","maxOut":8192,"in":0.15,"out":0.6,"cacheRead":0.075,"uptime":92,"noTools":1,"tps":17,"ttft":1449,"reqs":107},{"name":"Phala","maxOut":117964,"in":0.15,"out":0.6,"uptime":92.7,"tps":169,"ttft":1229,"reqs":5},{"name":"Together","maxOut":117964,"in":0.15,"out":0.6,"uptime":97.7,"tps":79,"ttft":322,"reqs":1349},{"name":"Groq","in":0.15,"out":0.6,"cacheRead":0.075,"uptime":100,"tps":314,"ttft":204,"reqs":57586},{"name":"Mara","maxOut":117964,"in":0.15,"out":0.75,"uptime":80.3,"tps":174,"ttft":1036,"reqs":104},{"name":"DeepInfra","quant":"fp8","maxOut":117964,"in":0.2,"out":0.95,"uptime":43,"noSchema":1,"tps":37,"ttft":464,"reqs":42062},{"name":"Cerebras","quant":"fp16","maxOut":40960,"in":0.35,"out":0.75,"cacheRead":0.35,"uptime":100,"tps":887,"ttft":218,"reqs":26877}],"scores":{"swebenchAgent":{"v":0.26,"as":"gpt-oss-120b","agent":"mini-SWE-agent","date":"2025-08-07","checked":true},"eqWriting":{"v":959.2,"as":"openai/gpt-oss-120b","version":"Creative Writing v3"},"eqLongform":{"v":35.8,"as":"openai/gpt-oss-120b","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.03,"regions":["eu","global","us"],"n":24,"tps":140,"ttft":1740,"tpsBest":887,"via":"Google","reqs":101505},"open":true,"rankable":true,"scored":3},{"id":"qwen/qwen3.6-35b-a3b","name":"Qwen: Qwen3.6 35B A3B","org":"Qwen","slug":"qwen3.6-35b-a3b","matchKey":"qwen3635ba3b","blurb":"Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per token.","created":1777260255,"cutoff":null,"hf":"Qwen/Qwen3.6-35B-A3B","ctx":262144,"maxOut":235929,"in":["text","image","video"],"out":["text"],"price":{"in":0.15,"out":1},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"fp4","maxOut":32768,"in":0.05,"out":0.7,"uptime":99.9,"tps":41,"ttft":4192,"reqs":5199},{"name":"AkashML","quant":"fp8","in":0.1,"out":0.9,"cacheRead":0.05,"uptime":99.9,"tps":58,"ttft":541,"reqs":17667},{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.1,"out":0.95,"uptime":98.2,"tps":20,"ttft":1552,"reqs":666},{"name":"Venice","quant":"fp8","ctx":256000,"maxOut":65536,"in":0.1,"out":1,"uptime":99.9,"noSchema":1,"tps":81,"ttft":627,"reqs":2610},{"name":"Parasail","quant":"fp8","in":0.15,"out":1,"cacheRead":0.05,"uptime":98.8,"tps":37,"ttft":480,"reqs":16213},{"name":"AtlasCloud","quant":"fp8","maxOut":65536,"in":0.186,"out":1.11375,"cacheRead":0.186,"uptime":99.8,"noSchema":1,"tps":115,"ttft":956,"reqs":11308},{"name":"Io Net","quant":"fp8","in":0.19,"out":0.99,"cacheRead":0.099,"uptime":89.7,"noSchema":1,"tps":40,"ttft":548,"reqs":529},{"name":"Phala","in":0.2,"out":1.27,"uptime":99.9,"tps":90,"ttft":1069,"reqs":993},{"name":"SiliconFlow","quant":"fp8","in":0.24,"out":1.8,"cacheRead":0.15,"uptime":99.1,"noTools":1,"tps":43,"ttft":1371,"reqs":550},{"name":"CoreWeave","quant":"fp8","in":0.25,"out":1.25,"cacheRead":0.25,"uptime":100,"noTools":1,"tps":82,"ttft":374,"reqs":1523}],"scores":{"frontierMath":{"v":0.20350877192982456,"as":"Qwen3.6 35B-A3B","effort":"","date":"2026-08-30","stderr":0.02389037106974561},"aime":{"v":0.8666666666666667,"as":"Qwen3.6 35B-A3B","effort":"","date":"2026-08-07","stderr":0.051247074319053836},"gpqa":{"v":0.8484848484848485,"as":"Qwen3.6 35B-A3B","effort":"","date":"2026-08-07","stderr":0.02554565042660364}},"perf":{"uptime":100,"n":10,"tps":58,"ttft":541,"tpsBest":115,"via":"AkashML","reqs":17667},"open":true,"rankable":true,"scored":3},{"id":"google/gemini-3.1-flash-lite","name":"Google: Gemini 3.1 Flash Lite","org":"Google","slug":"gemini-3.1-flash-lite","matchKey":"gemini31flashlite","blurb":"Gemini 3.1 Flash Lite is Google’s GA high-efficiency multimodal model optimized for low-latency, high-volume workloads.","created":1778168828,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.25,"out":1.5,"imageIn":2.5e-7,"audioIn":0.5,"webSearch":0.014,"alt":{"batch":{"in":0.125,"out":0.75}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Google","in":0.125,"out":0.75,"cacheRead":0.0125,"cacheWrite":0.041667,"uptime":8.7},{"name":"Google AI Studio","in":0.125,"out":0.75,"cacheRead":0.0125,"cacheWrite":0.041667,"uptime":100},{"name":"Google AI Studio","in":0.25,"out":1.5,"cacheRead":0.025,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"global","in":0.25,"out":1.5,"cacheRead":0.025,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"eu","in":0.275,"out":1.65,"cacheRead":0.0275,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"us","in":0.275,"out":1.65,"cacheRead":0.0275,"cacheWrite":0.083333,"uptime":100},{"name":"Google","in":0.45,"out":2.7,"cacheRead":0.045,"cacheWrite":0.15,"uptime":99.9,"tps":60,"ttft":526,"reqs":76,"region":"global"},{"name":"Google AI Studio","in":0.45,"out":2.7,"cacheRead":0.045,"cacheWrite":0.15,"uptime":99.9,"tps":92,"ttft":849,"reqs":62593}],"scores":{"frontierMath":{"v":0.2771929824561403,"as":"Gemini 3.1 Flash-Lite","effort":"","date":"2026-08-30","stderr":0.026560920735701255},"aime":{"v":0.8,"as":"Gemini 3.1 Flash-Lite","effort":"","date":"2026-08-06","stderr":0.06030226891555273},"gpqa":{"v":0.8181818181818182,"as":"Gemini 3.1 Flash-Lite","effort":"","date":"2026-08-06","stderr":0.027479603010538825}},"perf":{"uptime":100,"cache":0.0125,"regions":["eu","global","us"],"n":8,"tps":74,"ttft":912,"tpsBest":92,"via":"Google","reqs":226696},"open":false,"rankable":true,"scored":3},{"id":"qwen/qwen3.5-35b-a3b","name":"Qwen: Qwen3.5-35B-A3B","org":"Qwen","slug":"qwen3.5-35b-a3b","matchKey":"qwen3535ba3b","blurb":"The Qwen3.5 Series 35B-A3B is a native vision-language model designed with a hybrid architecture that integrates linear attention mechanisms and a sparse mixture-of-experts model, achieving higher inference efficiency.","created":1772053822,"cutoff":null,"hf":"Qwen/Qwen3.5-35B-A3B","ctx":262144,"maxOut":16384,"in":["text","image","video"],"out":["text"],"price":{"in":0.3125,"out":1.25},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"fp4","maxOut":32768,"in":0.08,"out":0.75,"uptime":99.9,"tps":53,"ttft":2578,"reqs":3596},{"name":"DeepInfra","quant":"fp8","maxOut":81920,"in":0.14,"out":1,"cacheRead":0.05,"uptime":99.9,"noTools":1,"tps":30,"ttft":252,"reqs":4200},{"name":"Parasail","quant":"fp8","maxOut":235929,"in":0.15,"out":1,"cacheRead":0.05,"uptime":99.9,"tps":42,"ttft":1329,"reqs":960},{"name":"Alibaba","maxOut":65536,"in":0.1625,"out":1.3,"uptime":99.8,"tps":43,"ttft":1914,"reqs":316},{"name":"AtlasCloud","quant":"fp8","maxOut":65536,"in":0.225,"out":1.8,"cacheRead":0.225,"uptime":99.2,"tps":75,"ttft":1990,"reqs":65},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.24,"out":1.8,"cacheRead":0.15,"uptime":65.5,"noTools":1,"tps":36,"ttft":1406,"reqs":70},{"name":"Venice","ctx":256000,"in":0.3125,"out":1.25,"cacheRead":0.15625,"uptime":100,"tps":21,"ttft":3802,"reqs":1111}],"scores":{"arenaChat":{"v":1394.9662806765355,"ci":4.3,"rank":156,"votes":29043,"as":"qwen3.5-35b-a3b","effort":""},"arenaWebdev":{"v":1250.14788605654,"ci":17.9,"rank":115,"votes":1535,"as":"qwen3.5-35b-a3b","effort":""},"eqLongform":{"v":44.5,"as":"Qwen/Qwen3.5-35B-A3B","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":7,"tps":30,"ttft":252,"tpsBest":75,"via":"DeepInfra","reqs":4200},"open":true,"rankable":true,"scored":3},{"id":"xiaomi/mimo-v2.5","name":"Xiaomi: MiMo-V2.5","org":"Xiaomi","slug":"mimo-v2.5","matchKey":"mimov25","blurb":"MiMo-V2.5 is a native omnimodal model by Xiaomi.","created":1776874269,"cutoff":null,"hf":"XiaomiMiMo/MiMo-V2.5","ctx":1050000,"maxOut":131072,"in":["text","audio","image","video"],"out":["text"],"price":{"in":0.14,"out":0.28},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"GMICloud","quant":"fp8","maxOut":945000,"in":0.119,"out":0.238,"cacheRead":0.00255,"uptime":94.8,"noSchema":1,"tps":22,"ttft":3448,"reqs":3966},{"name":"DeepInfra","quant":"fp8","ctx":262144,"in":0.133,"out":0.266,"cacheRead":0.00266,"uptime":98.7,"tps":21,"ttft":1065,"reqs":2416},{"name":"Xiaomi","quant":"fp8","ctx":1048576,"in":0.14,"out":0.28,"cacheRead":0.0028,"uptime":98.3,"noSchema":1,"tps":30,"ttft":2190,"reqs":241913},{"name":"StreamLake","ctx":1000000,"maxOut":128000,"in":0.168,"out":0.336,"cacheRead":0.00336,"uptime":93.2,"noTools":1,"tps":28,"ttft":1768,"reqs":635},{"name":"Novita","quant":"fp8","ctx":1048576,"in":0.168,"out":0.336,"cacheRead":0.0034,"uptime":98.7,"noSchema":1,"tps":29,"ttft":1993,"reqs":1621},{"name":"Venice","quant":"fp8","ctx":1000000,"maxOut":65536,"in":0.4,"out":2,"cacheRead":0.08,"uptime":98.5,"tps":19,"ttft":1913,"reqs":300}],"scores":{"arenaChat":{"v":1433.6144977411013,"ci":4.3,"rank":96,"votes":44466,"as":"mimo-v2.5","effort":""},"arenaWebdev":{"v":1437.1619826568476,"ci":6.2,"rank":63,"votes":14117,"as":"mimo-v2.5","effort":""},"arenaVision":{"v":1233.152310659354,"ci":6.3,"rank":58,"votes":24771,"as":"mimo-v2.5","effort":""}},"perf":{"uptime":98.7,"cache":0.00255,"n":6,"tps":30,"ttft":2190,"via":"Xiaomi","reqs":241913},"open":true,"rankable":true,"scored":3},{"id":"mistralai/mistral-nemo","name":"Mistral: Mistral Nemo","org":"Mistral","slug":"mistral-nemo","matchKey":"mistralnemo","blurb":"A 12B parameter model with a 128k token context length built by Mistral in collaboration with NVIDIA.","created":1721347200,"cutoff":"2024-04-30","hf":"mistralai/Mistral-Nemo-Instruct-2407","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.019,"out":0.03},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DekaLLM","quant":"fp8","maxOut":104857,"in":0.018,"out":0.03,"uptime":99.8,"noTools":1,"tps":31,"ttft":374,"reqs":40338},{"name":"DeepInfra","quant":"fp8","in":0.019,"out":0.03,"uptime":99.9,"tps":31,"ttft":499,"reqs":278357},{"name":"Parasail","quant":"fp8","maxOut":104857,"in":0.03,"out":0.03,"uptime":91.7,"noTools":1,"tps":54,"ttft":420,"reqs":2899},{"name":"Novita","quant":"fp8","ctx":60288,"maxOut":16000,"in":0.04,"out":0.17,"uptime":72.9,"noTools":1,"tps":37,"ttft":569,"reqs":2795},{"name":"Io Net","quant":"fp16","ctx":128000,"maxOut":102400,"in":0.04048,"out":0.1472,"cacheRead":0.02668,"uptime":98.4,"noSchema":1,"tps":6,"ttft":380,"reqs":8322}],"scores":{"bfcl":{"v":0.2763,"as":"Open-Mistral-Nemo-2407 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":879.1,"as":"mistralai/Mistral-Nemo-Instruct-2407","version":"Creative Writing v3"},"eqLongform":{"v":28.7,"as":"mistralai/Mistral-Nemo-Instruct-2407","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"n":5,"tps":31,"ttft":499,"tpsBest":54,"via":"DeepInfra","reqs":278357},"open":true,"prose":{"slop":26.406771123508896,"trigram":1.0112419577644303,"notXButY":0.22976260708034502,"repetition":10.087003806416531,"mattr":0.4901201735772239,"grade":8.624212003068205,"sentence":13.942616996990765,"words":262054,"samples":299,"as":"mistralai/mistral-nemo"},"rankable":true,"scored":3},{"id":"deepseek/deepseek-r1-0528","name":"DeepSeek: R1 0528","org":"DeepSeek","slug":"deepseek-r1-0528","matchKey":"deepseekr10528","blurb":"May 28th update to the [original DeepSeek R1](/deepseek/deepseek-r1) Performance on par with [OpenAI o1](/openai/o1), but open-sourced and with fully open reasoning tokens.","created":1748455170,"cutoff":"2025-03-31","hf":"deepseek-ai/DeepSeek-R1-0528","ctx":163840,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.5,"out":2.15},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","in":0.5,"out":2.15,"cacheRead":0.35,"uptime":99.9,"noTools":1,"tps":27,"ttft":755,"reqs":1789},{"name":"SiliconFlow","quant":"fp8","maxOut":147456,"in":0.5,"out":2.18,"uptime":99.4,"tps":20,"ttft":1238,"reqs":1310},{"name":"StreamLake","ctx":128000,"maxOut":32000,"in":0.571,"out":2.286,"uptime":98.3,"noTools":1,"tps":45,"ttft":3425,"reqs":387},{"name":"Novita","quant":"fp8","in":0.7,"out":2.5,"cacheRead":0.35,"uptime":100,"noSchema":1,"tps":25,"ttft":1054,"reqs":4}],"scores":{"arenaChat":{"v":1421.3686936624867,"ci":5.7,"rank":117,"votes":18091,"as":"deepseek-r1-0528","effort":""},"eqWriting":{"v":1418.8,"as":"deepseek-ai/DeepSeek-R1-0528","version":"Creative Writing v3"},"eqLongform":{"v":59.1,"as":"deepseek-ai/DeepSeek-R1-0528","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.35,"n":4,"tps":27,"ttft":755,"tpsBest":45,"via":"DeepInfra","reqs":1789},"open":true,"rankable":true,"scored":3},{"id":"x-ai/grok-4.20-multi-agent","name":"SpaceXAI: Grok 4.20 Multi-Agent","org":"SpaceXAI","slug":"grok-4.20-multi-agent","matchKey":"grok420multiagent","blurb":"Grok 4.20 Multi-Agent is a variant of SpaceXAI’s Grok 4.20 designed for collaborative, agent-based workflows.","created":1774979158,"cutoff":"2025-09-01","hf":null,"ctx":2000000,"maxOut":1800000,"in":["text","image","file"],"out":["text"],"price":{"in":1.25,"out":2.5,"webSearch":0.005},"caps":{"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"xAI","in":1.25,"out":2.5,"cacheRead":0.2,"uptime":87.7},{"name":"xAI","in":1.25,"out":2.5,"cacheRead":0.2,"uptime":76.2},{"name":"xAI","in":2.5,"out":5,"cacheRead":0.4},{"name":"xAI","in":2.5,"out":5,"cacheRead":0.4,"uptime":99,"tps":328,"ttft":9530,"reqs":23}],"scores":{"arenaChat":{"v":1470.1522547537338,"ci":3.8,"rank":40,"votes":60777,"as":"grok-4.20-multi-agent-beta-0309","effort":""},"arenaVision":{"v":1250.6196208726353,"ci":6.4,"rank":44,"votes":22925,"as":"grok-4.20-multi-agent-beta-0309","effort":""},"arenaSearch":{"v":1204.1098871642791,"ci":5.3,"rank":13,"votes":109553,"as":"grok-4.20-multi-agent-beta-0309","effort":""}},"perf":{"uptime":99,"cache":0.2,"n":4,"tps":328,"ttft":9530,"via":"xAI","reqs":23},"open":false,"rankable":true,"scored":3},{"id":"minimax/minimax-m2","name":"MiniMax: MiniMax M2","org":"MiniMax","slug":"minimax-m2","matchKey":"minimaxm2","blurb":"MiniMax-M2 is a compact, high-efficiency large language model optimized for end-to-end coding and agentic workflows.","created":1761252093,"cutoff":null,"hf":"MiniMaxAI/MiniMax-M2","ctx":204800,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.255,"out":1.02},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Minimax","quant":"fp8","in":0.255,"out":1.02,"uptime":99.5,"noSchema":1,"tps":78,"ttft":1446,"reqs":36},{"name":"Google","ctx":196608,"maxOut":176947,"in":0.3,"out":1.2,"uptime":100,"tps":113,"ttft":1030,"reqs":74},{"name":"Novita","quant":"fp8","in":0.3,"out":1.2,"cacheRead":0.03,"uptime":100,"noSchema":1,"tps":61,"ttft":1414,"reqs":31}],"scores":{"arenaWebdev":{"v":1297.7787549249838,"ci":11,"rank":108,"votes":6539,"as":"minimax-m2","effort":""},"swebenchAgent":{"v":0.61,"as":"MiniMax M2","agent":"mini-SWE-agent","date":"2025-11-24","checked":true},"eqLongform":{"v":46.4,"as":"minimax/minimax-m2","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":3,"tps":113,"ttft":1030,"via":"Google","reqs":74},"open":true,"rankable":true,"scored":3},{"id":"minimax/minimax-m2.1","name":"MiniMax: MiniMax M2.1","org":"MiniMax","slug":"minimax-m2.1","matchKey":"minimaxm21","blurb":"MiniMax-M2.1 is a lightweight, state-of-the-art large language model optimized for coding, agentic workflows, and modern application development.","created":1766454997,"cutoff":null,"hf":"MiniMaxAI/MiniMax-M2.1","ctx":204800,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"fp8","in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.2,"tps":41,"ttft":1566,"reqs":19},{"name":"Minimax","quant":"fp8","in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.1},{"name":"Minimax","quant":"fp8","in":0.3,"out":2.4,"cacheRead":0.03,"uptime":99.4,"tps":60,"ttft":1942,"reqs":18}],"scores":{"arenaChat":{"v":1383.788625927177,"ci":5.3,"rank":170,"votes":16622,"as":"minimax-m2.1-preview","effort":""},"arenaWebdev":{"v":1387.4824723418183,"ci":9.7,"rank":82,"votes":7281,"as":"minimax-m2.1-preview","effort":""},"eqLongform":{"v":57.8,"as":"MiniMaxAI/MiniMax-M2.1","version":"Longform Writing v1.11"}},"perf":{"uptime":99.4,"cache":0.03,"n":3,"tps":41,"ttft":1566,"tpsBest":60,"via":"Novita","reqs":19},"open":true,"rankable":true,"scored":3},{"id":"mistralai/mistral-small-2603","name":"Mistral: Mistral Small 4","org":"Mistral","slug":"mistral-small-2603","matchKey":"mistralsmall2603","blurb":"Mistral Small 4 is the next major release in the Mistral Small family, unifying the capabilities of several flagship Mistral models into a single system.","created":1773695685,"cutoff":null,"hf":"mistralai/Mistral-Small-4-119B-2603","ctx":262144,"maxOut":209715,"in":["text","image"],"out":["text"],"price":{"in":0.15,"out":0.6,"alt":{"batch":{"in":0.075,"out":0.3}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","none"]},"moderated":false,"providers":[{"name":"Mistral","in":0.15,"out":0.6,"cacheRead":0.015,"uptime":99.9},{"name":"Mistral","in":0.15,"out":0.6,"cacheRead":0.015,"uptime":99.9},{"name":"Mistral","region":"eu","in":0.165,"out":0.66,"cacheRead":0.0165,"uptime":100,"tps":38,"ttft":548,"reqs":11}],"scores":{"arenaChat":{"v":1356.7230679844458,"ci":5.2,"rank":196,"votes":17391,"as":"mistral-small-2506","effort":""},"arenaVision":{"v":1139.6518569576085,"ci":9.1,"rank":110,"votes":10869,"as":"mistral-small-2506","effort":""},"eqLongform":{"v":43.2,"as":"mistralai/Mistral-Small-4-119B-2603","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.015,"regions":["eu"],"n":3,"tps":106,"ttft":324,"via":"Mistral","reqs":2513},"open":true,"rankable":true,"scored":3},{"id":"mistralai/mistral-small-3.2-24b-instruct","name":"Mistral: Mistral Small 3.2 24B","org":"Mistral","slug":"mistral-small-3.2-24b-instruct","matchKey":"mistralsmall3224binstruct","blurb":"Mistral-Small-3.2-24B-Instruct-2506 is an updated 24B parameter model from Mistral optimized for instruction following, repetition reduction, and improved function calling.","created":1750443016,"cutoff":"2023-10-31","hf":"mistralai/Mistral-Small-3.2-24B-Instruct-2506","ctx":256000,"maxOut":16384,"in":["image","text"],"out":["text"],"price":{"in":0.09375,"out":0.25},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","ctx":128000,"in":0.075,"out":0.2,"uptime":99.6,"tps":35,"ttft":706,"reqs":3610},{"name":"Parasail","quant":"bf16","ctx":131072,"maxOut":32768,"in":0.09,"out":0.3,"cacheRead":0.05,"uptime":99.8,"noTools":1,"tps":45,"ttft":538,"reqs":1465},{"name":"Venice","quant":"fp8","in":0.09375,"out":0.25,"uptime":99.2,"tps":28,"ttft":799,"reqs":15214}],"scores":{"bfcl":{"v":0.3715,"as":"Mistral-small-2506 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1252.9,"as":"mistralai/Mistral-Small-3.2-24B-Instruct-2506","version":"Creative Writing v3"},"eqLongform":{"v":41.6,"as":"mistralai/Mistral-Small-3.2-24B-Instruct-2506","version":"Longform Writing v1.11"}},"perf":{"uptime":99.8,"n":3,"tps":28,"ttft":799,"tpsBest":45,"via":"Venice","reqs":15214},"open":true,"rankable":true,"scored":3},{"id":"openai/gpt-5.3-codex","name":"OpenAI: GPT-5.3-Codex","org":"OpenAI","slug":"gpt-5.3-codex","matchKey":"gpt53codex","blurb":"GPT-5.3-Codex is OpenAI’s most advanced agentic coding model, combining the frontier software engineering performance of GPT-5.2-Codex with the broader reasoning and professional knowledge capabilities of GPT-5.2.","created":1771959164,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":1.75,"out":14,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"Azure","in":1.75,"out":14,"cacheRead":0.175,"uptime":100,"tps":42,"ttft":3959,"reqs":122},{"name":"OpenAI","in":1.75,"out":14,"cacheRead":0.175,"uptime":100},{"name":"OpenAI","in":3.5,"out":28,"cacheRead":0.35,"tps":88,"ttft":2065,"reqs":583}],"scores":{"arenaWebdev":{"v":1408.7914258973374,"ci":13.9,"rank":72,"votes":2514,"as":"gpt-5.3-codex (codex-harness)","effort":""},"arenaImageCode":{"v":1444.4876947509435,"ci":14.3,"rank":39,"votes":2539,"as":"gpt-5.3-codex (codex-harness)","effort":""},"swebench":{"v":0.7479338842975206,"as":"GPT-5.3 Codex (high)","effort":"high","date":"2026-02-25","stderr":0.019756739501473306}},"perf":{"uptime":100,"cache":0.175,"n":3,"tps":88,"ttft":2065,"via":"OpenAI","reqs":583},"open":false,"rankable":true,"scored":3},{"id":"openai/gpt-5.4-pro","name":"OpenAI: GPT-5.4 Pro","org":"OpenAI","slug":"gpt-5.4-pro","matchKey":"gpt54pro","blurb":"GPT-5.4 Pro is OpenAI's most advanced model, building on GPT-5.4's unified architecture with enhanced reasoning capabilities for complex, high-stakes tasks.","created":1772734366,"cutoff":null,"hf":null,"ctx":1050000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":30,"out":180,"webSearch":0.01,"alt":{"batch":{"in":15,"out":90}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium"]},"moderated":true,"providers":[{"name":"OpenAI","in":15,"out":90},{"name":"Azure","in":30,"out":180,"uptime":100,"tps":2,"ttft":5127,"reqs":3},{"name":"OpenAI","in":30,"out":180,"uptime":99.9,"tps":2,"ttft":5314,"reqs":25}],"scores":{"frontierMath":{"v":0.8245614035087719,"as":"GPT-5.4 Pro (xhigh)","effort":"xhigh","date":"2026-06-13","stderr":0.022569134425265196},"gpqa":{"v":0.946,"as":"GPT-5.4 Pro (xhigh)","effort":"xhigh","date":"2026-03-20","stderr":0.016},"simpleqa":{"v":0.463,"as":"GPT-5.4 Pro (xhigh)","effort":"xhigh","date":"2026-08-27","stderr":0.015775927227262426}},"perf":{"uptime":100,"n":3,"tps":2,"ttft":5314,"via":"OpenAI","reqs":25},"open":false,"rankable":true,"scored":3},{"id":"z-ai/glm-4.7-flash","name":"Z.ai: GLM 4.7 Flash","org":"Z.ai","slug":"glm-4.7-flash","matchKey":"glm47flash","blurb":"As a 30B-class SOTA model, GLM-4.7-Flash offers a new option that balances performance and efficiency.","created":1768833913,"cutoff":null,"hf":"zai-org/GLM-4.7-Flash","ctx":200000,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.0605,"out":0.4},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Venice","quant":"fp8","ctx":128000,"maxOut":16384,"in":0.06,"out":0.4,"cacheRead":0.01,"uptime":97.8,"tps":15,"ttft":1706,"reqs":640},{"name":"Cloudflare","ctx":131072,"in":0.0605,"out":0.4,"uptime":99,"tps":23,"ttft":347,"reqs":9152},{"name":"Novita","quant":"bf16","maxOut":128000,"in":0.07,"out":0.4,"cacheRead":0.01,"uptime":60.2,"noSchema":1,"tps":42,"ttft":1481,"reqs":263}],"scores":{"arenaChat":{"v":1366.264649011693,"ci":5.8,"rank":186,"votes":11491,"as":"glm-4.7-flash","effort":""},"eqWriting":{"v":1122.6,"as":"zai-org/GLM-4.7-Flash","version":"Creative Writing v3"},"eqLongform":{"v":47.8,"as":"zai-org/GLM-4.7-Flash","version":"Longform Writing v1.11"}},"perf":{"uptime":99,"cache":0.01,"n":3,"tps":23,"ttft":347,"tpsBest":42,"via":"Cloudflare","reqs":9152},"open":true,"rankable":true,"scored":3},{"id":"moonshotai/kimi-k2-thinking","name":"MoonshotAI: Kimi K2 Thinking","org":"MoonshotAI","slug":"kimi-k2-thinking","matchKey":"kimik2thinking","blurb":"Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning.","created":1762440622,"cutoff":null,"hf":"moonshotai/Kimi-K2-Thinking","ctx":262144,"maxOut":98304,"in":["text"],"out":["text"],"price":{"in":0.6,"out":2.5},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Google","maxOut":235929,"in":0.6,"out":2.5,"uptime":100,"tps":129,"ttft":754,"reqs":130},{"name":"Novita","quant":"bf16","in":0.6,"out":2.5,"cacheRead":0.15,"uptime":99,"tps":43,"ttft":1698,"reqs":165}],"scores":{"swebenchAgent":{"v":0.634,"as":"Kimi K2 Thinking","agent":"mini-SWE-agent","date":"2025-12-10","checked":true},"eqWriting":{"v":1627.9,"as":"moonshotai/Kimi-K2-Thinking","version":"Creative Writing v3"},"eqLongform":{"v":69,"as":"moonshotai/Kimi-K2-Thinking","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":2,"tps":43,"ttft":1698,"tpsBest":129,"via":"Novita","reqs":165},"open":true,"rankable":true,"scored":3},{"id":"openai/gpt-4","name":"OpenAI: GPT-4","org":"OpenAI","slug":"gpt-4","matchKey":"gpt4","blurb":"OpenAI's flagship model, GPT-4 is a large-scale multimodal language model capable of solving difficult problems with greater accuracy than previous models due to its broader general knowledge and advanced reasoning...","created":1685232000,"cutoff":"2021-09-30","hf":null,"ctx":8191,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":30,"out":60},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":30,"out":60,"uptime":99.9,"noSchema":1,"tps":18,"ttft":593,"reqs":23},{"name":"OpenAI","in":30,"out":60,"uptime":99.9,"tps":11,"ttft":568,"reqs":16}],"scores":{"aime":{"v":0.011111111111111112,"as":"GPT-4 (Jun 2023)","effort":"","date":"2025-10-23","stderr":0.005362818328685681},"gpqa":{"v":0.3573232323232323,"as":"GPT-4 (Mar 2023)","effort":"","date":"2025-10-23","stderr":0.02363099375599328},"swebenchAgent":{"v":0.22399999999999998,"as":"GPT-4 (1106)","agent":"SWE-agent","date":"2024-04-02","checked":true}},"perf":{"uptime":99.9,"n":2,"tps":18,"ttft":593,"via":"Azure","reqs":23},"open":false,"rankable":true,"scored":3},{"id":"openai/gpt-5.5-pro","name":"OpenAI: GPT-5.5 Pro","org":"OpenAI","slug":"gpt-5.5-pro","matchKey":"gpt55pro","blurb":"GPT-5.5 Pro is OpenAI’s high-capability model optimized for deep reasoning and accuracy on complex, high-stakes workloads.","created":1777051896,"cutoff":"2025-12-01","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":30,"out":180,"webSearch":0.01,"alt":{"batch":{"in":15,"out":90}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium"]},"moderated":true,"providers":[{"name":"OpenAI","in":15,"out":90},{"name":"OpenAI","in":30,"out":180,"uptime":100,"tps":5,"ttft":3162,"reqs":324}],"scores":{"frontierMath":{"v":0.8771929824561403,"as":"GPT-5.5 Pro (xhigh)","effort":"xhigh","date":"2026-06-12","stderr":0.019476010341530077},"aime":{"v":1,"as":"GPT-5.5 Pro (xhigh)","effort":"xhigh","date":"2026-04-24","stderr":0},"gpqa":{"v":0.9392135642135643,"as":"GPT-5.5 Pro (xhigh)","effort":"xhigh","date":"2026-04-24","stderr":0.015832823810775415}},"perf":{"uptime":100,"n":2,"tps":5,"ttft":3162,"via":"OpenAI","reqs":324},"open":false,"rankable":true,"scored":3},{"id":"google/gemma-3-12b-it","name":"Google: Gemma 3 12B","org":"Google","slug":"gemma-3-12b-it","matchKey":"gemma312bit","blurb":"Gemma 3 introduces multimodality, supporting vision-language input and text outputs.","created":1741902625,"cutoff":"2024-08-31","hf":"google/gemma-3-12b-it","ctx":131072,"maxOut":16384,"in":["text","image"],"out":["text"],"price":{"in":0.05,"out":0.15},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":0.05,"out":0.15,"uptime":99.9,"tps":6,"ttft":1159,"reqs":25303}],"scores":{"bfcl":{"v":0.3043,"as":"Gemma-3-12b-it (Prompt)","effort":"","version":"V4","agent":"BFCL V4 · prompt-based tool calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1123.4,"as":"google/gemma-3-12b-it","version":"Creative Writing v3"},"eqLongform":{"v":41.8,"as":"google/gemma-3-12b-it","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"n":1,"tps":6,"ttft":1159,"via":"DeepInfra","reqs":25303},"open":true,"prose":{"slop":36.3532116414076,"trigram":0.9978664402630838,"notXButY":0.3077912133055163,"repetition":21.23183797270192,"mattr":0.5661301055826549,"grade":10.70764266252445,"sentence":14.33640445958501,"words":362774,"samples":300,"as":"google/gemma-3-12b-it"},"rankable":true,"scored":3},{"id":"google/gemma-3-4b-it","name":"Google: Gemma 3 4B","org":"Google","slug":"gemma-3-4b-it","matchKey":"gemma34bit","blurb":"Gemma 3 introduces multimodality, supporting vision-language input and text outputs.","created":1741905510,"cutoff":"2024-08-31","hf":"google/gemma-3-4b-it","ctx":131072,"maxOut":16384,"in":["text","image"],"out":["text"],"price":{"in":0.05,"out":0.1},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":0.05,"out":0.1,"uptime":100,"tps":21,"ttft":1656,"reqs":1263}],"scores":{"bfcl":{"v":0.1962,"as":"Gemma-3-4b-it (Prompt)","effort":"","version":"V4","agent":"BFCL V4 · prompt-based tool calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1065.9,"as":"google/gemma-3-4b-it","version":"Creative Writing v3"},"eqLongform":{"v":34.4,"as":"google/gemma-3-4b-it","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":1,"tps":21,"ttft":1656,"via":"DeepInfra","reqs":1263},"open":true,"prose":{"slop":40.44117647058824,"trigram":0.585844093735055,"notXButY":0.8114468877690987,"repetition":12.446411284746231,"mattr":0.5706022148805723,"grade":8.439864657835855,"sentence":13.818174513900049,"words":167280,"samples":150,"as":"google/gemma-3-4b-it"},"rankable":true,"scored":3},{"id":"openai/gpt-3.5-turbo","name":"OpenAI: GPT-3.5 Turbo","org":"OpenAI","slug":"gpt-3.5-turbo","matchKey":"gpt35turbo","blurb":"GPT-3.5 Turbo is OpenAI's fastest model.","created":1685232000,"cutoff":"2021-09-30","hf":null,"ctx":16385,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":0.5,"out":1.5,"alt":{"batch":{"in":0.25,"out":0.75}}},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":0.5,"out":1.5,"uptime":100,"tps":49,"ttft":618,"reqs":775}],"scores":{"frontierMath":{"v":0,"as":"GPT-3.5 Turbo (Jan 2024)","effort":"","date":"2026-08-27","stderr":0},"aime":{"v":0.022222222222222223,"as":"GPT-3.5 Turbo (Jan 2024)","effort":"","date":"2026-08-07","stderr":0.022222222222222223},"gpqa":{"v":0.2803030303030303,"as":"GPT-3.5 Turbo (Nov 2023)","effort":"","date":"2025-01-27","stderr":0.018720372561004013}},"perf":{"uptime":100,"n":1,"tps":49,"ttft":618,"via":"OpenAI","reqs":775},"open":false,"rankable":true,"scored":3},{"id":"qwen/qwen3.5-plus-20260420","name":"Qwen: Qwen3.5 Plus 2026-04-20","org":"Qwen","slug":"qwen3.5-plus-20260420","matchKey":"qwen35plus20260420","blurb":"Qwen3.5 Plus (April 2026) is a large-scale multimodal language model from Alibaba.","created":1777261368,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.3,"out":1.8},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.3,"out":1.8,"cacheWrite":0.375,"uptime":100,"tps":47,"ttft":1085,"reqs":472}],"scores":{"aime":{"v":0.8666666666666667,"as":"Qwen3.5 Plus","effort":"","date":"2026-08-07","stderr":0.051247074319053836},"gpqa":{"v":0.8484848484848485,"as":"Qwen3.5 Plus","effort":"","date":"2026-08-07","stderr":0.025545650426603637},"simpleqa":{"v":0.2535070140280561,"as":"Qwen3.5 Plus","effort":"","date":"2026-08-27","stderr":0.013777178222916077}},"perf":{"uptime":100,"n":1,"tps":47,"ttft":1085,"via":"Alibaba","reqs":472},"open":false,"rankable":true,"scored":3},{"id":"qwen/qwen3.7-flash","name":"Qwen: Qwen3.7 Flash","org":"Qwen","slug":"qwen3.7-flash","matchKey":"qwen37flash","blurb":"Qwen3.7 Flash is a vision-language reasoning model from Alibaba.","created":1785190561,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.03,"out":0.13},"caps":{"tools":true,"json":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.03,"out":0.13,"cacheRead":0.006,"cacheWrite":0.038,"uptime":100,"tps":57,"ttft":656,"reqs":256718}],"scores":{"frontierMath":{"v":0.19298245614035087,"as":"Qwen3.7 Flash","effort":"","date":"2026-08-29","stderr":0.02341753795065531},"aime":{"v":0.8666666666666667,"as":"Qwen3.7 Flash","effort":"","date":"2026-08-07","stderr":0.051247074319053836},"gpqa":{"v":0.8232323232323232,"as":"Qwen3.7 Flash","effort":"","date":"2026-08-07","stderr":0.027178752639044908}},"perf":{"uptime":100,"cache":0.006,"n":1,"tps":57,"ttft":656,"via":"Alibaba","reqs":256718},"open":false,"rankable":true,"scored":3},{"id":"openai/gpt-oss-20b","name":"OpenAI: gpt-oss-20b","org":"OpenAI","slug":"gpt-oss-20b","matchKey":"gptoss20b","blurb":"gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.","created":1754414229,"cutoff":"2024-06-30","hf":"openai/gpt-oss-20b","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.03,"out":0.13},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"fp8","maxOut":32768,"in":0.018,"out":0.09,"uptime":99.7,"tps":30,"ttft":3182,"reqs":20787},{"name":"AkashML","quant":"fp4","in":0.02,"out":0.1,"uptime":99.4,"tps":25,"ttft":1031,"reqs":6673},{"name":"DekaLLM","quant":"bf16","in":0.029,"out":0.14,"uptime":99.5,"tps":20,"ttft":1010,"reqs":3612},{"name":"CoreWeave","quant":"fp4","in":0.03,"out":0.13,"cacheRead":0.03,"uptime":100,"tps":166,"ttft":124,"reqs":20092},{"name":"DeepInfra","quant":"bf16","in":0.03,"out":0.14,"uptime":99.9,"tps":93,"ttft":307,"reqs":11715},{"name":"Parasail","quant":"fp4","in":0.03,"out":0.15,"cacheRead":0.02,"uptime":99.7,"tps":56,"ttft":816,"reqs":26808},{"name":"Novita","quant":"fp4","maxOut":32768,"in":0.04,"out":0.15,"uptime":99.9,"noTools":1,"tps":149,"ttft":678,"reqs":1498},{"name":"SiliconFlow","quant":"fp8","maxOut":8192,"in":0.04,"out":0.18,"uptime":96.3,"noTools":1,"tps":43,"ttft":1147,"reqs":661},{"name":"Together","in":0.05,"out":0.2,"noTools":1},{"name":"Amazon Bedrock","region":"eu","in":0.07,"out":0.15,"uptime":100,"noSchema":1},{"name":"Amazon Bedrock","in":0.07,"out":0.15,"uptime":97.7,"noSchema":1,"tps":176,"ttft":543,"reqs":3371,"region":"eu"},{"name":"Google","region":"us","maxOut":32768,"in":0.07,"out":0.25,"uptime":98.8,"noTools":1,"tps":169,"ttft":732,"reqs":1597},{"name":"Groq","maxOut":65536,"in":0.075,"out":0.3,"cacheRead":0.0375,"uptime":99.8,"tps":293,"ttft":889,"reqs":24382}],"scores":{"eqWriting":{"v":663.7,"as":"openai/gpt-oss-20b","version":"Creative Writing v3"},"eqLongform":{"v":21.2,"as":"openai/gpt-oss-20b","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"regions":["eu","us"],"n":13,"tps":56,"ttft":816,"tpsBest":293,"via":"Parasail","reqs":26808},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3-235b-a22b-2507","name":"Qwen: Qwen3 235B A22B Instruct 2507","org":"Qwen","slug":"qwen3-235b-a22b-2507","matchKey":"qwen3235ba22b2507","blurb":"Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass.","created":1753119555,"cutoff":"2025-06-30","hf":"Qwen/Qwen3-235B-A22B-Instruct-2507","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.0875,"out":0.35},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"GMICloud","quant":"fp8","in":0.0875,"out":0.35,"cacheRead":0.0175,"uptime":98.6,"tps":23,"ttft":1278,"reqs":26538},{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.09,"out":0.55,"uptime":94.9,"tps":4,"ttft":5719,"reqs":1859},{"name":"Novita","quant":"fp8","ctx":131072,"maxOut":16384,"in":0.09,"out":0.58,"uptime":96.8,"tps":24,"ttft":893,"reqs":8075},{"name":"Parasail","quant":"fp8","ctx":131072,"maxOut":117964,"in":0.14,"out":0.8,"cacheRead":0.05,"uptime":99.7,"tps":19,"ttft":649,"reqs":15708},{"name":"Alibaba","ctx":131072,"maxOut":32768,"in":0.1495,"out":0.598,"uptime":100,"noSchema":1,"tps":44,"ttft":481,"reqs":10892},{"name":"Venice","quant":"fp8","ctx":128000,"maxOut":16384,"in":0.15,"out":0.75,"uptime":95.6,"tps":11,"ttft":1055,"reqs":5943},{"name":"Nebius","quant":"fp8","in":0.2,"out":0.6,"uptime":92.3,"tps":20,"ttft":646,"reqs":20535},{"name":"StreamLake","ctx":128000,"maxOut":32000,"in":0.21,"out":0.84,"uptime":98.9,"tps":27,"ttft":846,"reqs":7207},{"name":"Google","region":"us","maxOut":16384,"in":0.22,"out":0.88,"uptime":99.9},{"name":"Google","region":"us","maxOut":16384,"in":0.25,"out":1,"uptime":99.9,"noTools":1,"noSchema":1,"tps":16,"ttft":739,"reqs":5569}],"scores":{"bfcl":{"v":0.4799,"as":"Qwen3-235B-A22B-Instruct-2507 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqLongform":{"v":57.4,"as":"Qwen/Qwen3-235B-A22B-Instruct-2507","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.0175,"regions":["us"],"n":10,"tps":23,"ttft":1278,"tpsBest":44,"via":"GMICloud","reqs":26538},"open":true,"rankable":true,"scored":2},{"id":"minimax/minimax-m2.7","name":"MiniMax: MiniMax M2.7","org":"MiniMax","slug":"minimax-m2.7","matchKey":"minimaxm27","blurb":"MiniMax-M2.7 is a next-generation large language model designed for autonomous, real-world productivity and continuous improvement.","created":1773836697,"cutoff":null,"hf":"MiniMaxAI/MiniMax-M2.7","ctx":204800,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"GMICloud","quant":"fp8","ctx":196608,"maxOut":176947,"in":0.21,"out":0.84,"cacheRead":0.042,"uptime":99.6,"noSchema":1,"tps":45,"ttft":1634,"reqs":1177},{"name":"Mara","ctx":196608,"maxOut":176947,"in":0.24,"out":0.96,"cacheRead":0.024,"uptime":81.8,"tps":123,"ttft":5644,"reqs":1305},{"name":"Novita","quant":"fp8","in":0.27,"out":1.08,"cacheRead":0.054,"uptime":99.6,"noSchema":1,"tps":58,"ttft":1098,"reqs":2420},{"name":"AtlasCloud","quant":"fp8","ctx":196608,"maxOut":176947,"in":0.3,"out":1.2,"cacheRead":0.06,"uptime":98.6,"noSchema":1,"tps":26,"ttft":2308,"reqs":1730},{"name":"Minimax","quant":"fp8","in":0.3,"out":1.2,"cacheRead":0.06,"uptime":99.6,"noSchema":1},{"name":"DeepInfra","quant":"fp8","ctx":196608,"in":0.38,"out":1.7,"cacheRead":0.07,"uptime":96.2,"noSchema":1,"tps":63,"ttft":1744,"reqs":77},{"name":"Groq","ctx":196608,"in":0.6,"out":1.8,"uptime":99.4,"noSchema":1,"tps":244,"ttft":271,"reqs":285},{"name":"SambaNova","ctx":196608,"maxOut":176947,"in":0.6,"out":2.4,"cacheRead":0.06,"uptime":98.6,"noSchema":1,"tps":38,"ttft":2235,"reqs":43},{"name":"Minimax","quant":"fp8","in":0.6,"out":2.4,"cacheRead":0.06,"uptime":93.1,"noSchema":1,"tps":58,"ttft":1926,"reqs":3241}],"scores":{"arenaChat":{"v":1415.173845342476,"ci":3.6,"rank":126,"votes":69169,"as":"minimax-m2.7","effort":""},"arenaWebdev":{"v":1397.5205307992737,"ci":5.9,"rank":77,"votes":15194,"as":"minimax-m2.7","effort":""}},"perf":{"uptime":99.6,"cache":0.042,"n":9,"tps":58,"ttft":1926,"tpsBest":244,"via":"Minimax","reqs":3241},"open":true,"rankable":true,"scored":2},{"id":"deepseek/deepseek-chat-v3.1","name":"DeepSeek: DeepSeek V3.1","org":"DeepSeek","slug":"deepseek-chat-v3.1","matchKey":"deepseekchatv31","blurb":"DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates.","created":1755779628,"cutoff":"2025-03-31","hf":"deepseek-ai/DeepSeek-V3.1","ctx":163840,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.25,"out":0.95},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","in":0.25,"out":0.95,"cacheRead":0.13,"uptime":99.9,"tps":11,"ttft":795,"reqs":7711},{"name":"SiliconFlow","quant":"fp8","maxOut":147456,"in":0.27,"out":1,"uptime":95.4,"tps":17,"ttft":2037,"reqs":1154},{"name":"Novita","quant":"fp8","ctx":131072,"in":0.27,"out":1,"cacheRead":0.135,"uptime":100,"tps":19,"ttft":1751,"reqs":2242},{"name":"AtlasCloud","quant":"fp8","ctx":131072,"maxOut":65536,"in":0.3,"out":0.95,"cacheRead":0.13,"uptime":99.2,"noTools":1,"tps":32,"ttft":1599,"reqs":978},{"name":"CoreWeave","quant":"fp8","ctx":161000,"maxOut":144900,"in":0.55,"out":1.65,"cacheRead":0.55,"uptime":100,"tps":49,"ttft":343,"reqs":4885},{"name":"Mara","ctx":131072,"maxOut":117964,"in":0.6,"out":1.7,"uptime":93.6,"noTools":1,"tps":92,"ttft":641,"reqs":2225},{"name":"Google","region":"us","in":0.6,"out":1.7,"uptime":0},{"name":"SambaNova","quant":"fp8","ctx":131072,"maxOut":7168,"in":0.65,"out":1.5,"uptime":99.7,"noSchema":1,"tps":57,"ttft":1158,"reqs":448}],"scores":{"eqWriting":{"v":1433.2,"as":"deepseek-ai/DeepSeek-V3.1","version":"Creative Writing v3"},"eqLongform":{"v":62.3,"as":"deepseek-ai/DeepSeek-V3.1","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.13,"regions":["us"],"n":8,"tps":11,"ttft":795,"tpsBest":92,"via":"DeepInfra","reqs":7711},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3.5-9b","name":"Qwen: Qwen3.5-9B","org":"Qwen","slug":"qwen3.5-9b","matchKey":"qwen359b","blurb":"Qwen3.5-9B is a multimodal foundation model from the Qwen3.5 family, designed to deliver strong reasoning, coding, and visual understanding in an efficient 9B-parameter architecture.","created":1773152396,"cutoff":null,"hf":"Qwen/Qwen3.5-9B","ctx":262144,"maxOut":32768,"in":["text","image","video"],"out":["text"],"price":{"in":0.1,"out":0.15},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"fp4","maxOut":65536,"in":0.08,"out":0.13,"uptime":99.9,"tps":27,"ttft":1492,"reqs":838},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.1,"out":0.15,"uptime":96.9,"tps":36,"ttft":810,"reqs":4496},{"name":"DeepInfra","quant":"bf16","maxOut":81920,"in":0.1,"out":0.15,"uptime":99.9,"tps":10,"ttft":1099,"reqs":5289},{"name":"Venice","quant":"fp8","ctx":256000,"in":0.1,"out":0.15,"uptime":99.9,"tps":80,"ttft":556,"reqs":5066},{"name":"Parasail","quant":"bf16","maxOut":235929,"in":0.1,"out":0.25,"uptime":99.8,"noTools":1,"tps":40,"ttft":449,"reqs":761},{"name":"Together","maxOut":235929,"in":0.17,"out":0.25,"uptime":97.4,"tps":42,"ttft":393,"reqs":2824}],"scores":{"aime":{"v":0.6166666666666667,"as":"Qwen3.5 9B","effort":"","date":"2026-08-27","stderr":0.06003996985174109},"gpqa":{"v":0.7897727272727273,"as":"Qwen3.5 9B","effort":"","date":"2026-08-27","stderr":0.02365568030958479}},"perf":{"uptime":99.9,"n":6,"tps":10,"ttft":1099,"tpsBest":80,"via":"DeepInfra","reqs":5289},"open":true,"rankable":true,"scored":2},{"id":"tencent/hy3","name":"Tencent: Hy3","org":"Tencent","slug":"hy3","matchKey":"hy3","blurb":"Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use.","created":1783344048,"cutoff":null,"hf":"tencent/Hy3","ctx":262144,"maxOut":128000,"in":["text"],"out":["text"],"price":{"in":0.132,"out":0.528},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","low","none"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","maxOut":131072,"in":0.13,"out":0.53,"cacheRead":0.033,"uptime":99.9,"tps":35,"ttft":1489,"reqs":1180},{"name":"Tencent","quant":"fp8","in":0.132,"out":0.528,"cacheRead":0.033,"uptime":99.7,"noSchema":1,"tps":33,"ttft":4462,"reqs":223819},{"name":"Novita","maxOut":235929,"in":0.14,"out":0.58,"cacheRead":0.035,"uptime":99.4,"tps":30,"ttft":3110,"reqs":1695},{"name":"GMICloud","quant":"bf16","maxOut":235929,"in":0.14,"out":0.58,"cacheRead":0.035,"uptime":99.5,"noSchema":1,"tps":6,"ttft":2227,"reqs":174},{"name":"Phala","maxOut":235929,"in":0.15,"out":0.64,"cacheRead":0.04,"uptime":99.8,"tps":8,"ttft":1560,"reqs":142},{"name":"AtlasCloud","quant":"fp8","maxOut":131072,"in":0.2,"out":0.8,"cacheRead":0.05,"uptime":99.9,"tps":13,"ttft":1931,"reqs":437}],"scores":{"arenaChat":{"v":1455.7789602016828,"ci":7.2,"rank":60,"votes":8047,"as":"hy3","effort":""},"arenaWebdev":{"v":1512.6396407346306,"ci":10.6,"rank":40,"votes":3586,"as":"hy3","effort":""}},"perf":{"uptime":99.9,"cache":0.033,"n":6,"tps":33,"ttft":4462,"tpsBest":35,"via":"Tencent","reqs":223819},"open":true,"rankable":true,"scored":2},{"id":"meta-llama/llama-3.1-8b-instruct","name":"Meta: Llama 3.1 8B Instruct","org":"Meta","slug":"llama-3.1-8b-instruct","matchKey":"llama318binstruct","blurb":"Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors.","created":1721692800,"cutoff":"2023-12-31","hf":"meta-llama/Meta-Llama-3.1-8B-Instruct","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.05,"out":0.08},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.02,"out":0.04,"uptime":99.9,"noTools":1,"noSchema":1,"tps":22,"ttft":860,"reqs":13886},{"name":"Novita","quant":"fp8","ctx":16384,"maxOut":14745,"in":0.02,"out":0.05,"uptime":99.3,"noTools":1,"noSchema":1,"tps":58,"ttft":528,"reqs":26606},{"name":"Groq","in":0.05,"out":0.08,"cacheRead":0.025,"uptime":99.9,"noSchema":1,"tps":115,"ttft":221,"reqs":52514},{"name":"Cloudflare","quant":"fp8","ctx":32000,"maxOut":28800,"in":0.152,"out":0.287,"uptime":98.5,"noTools":1,"noSchema":1,"tps":12,"ttft":559,"reqs":2518},{"name":"CoreWeave","quant":"bf16","in":0.22,"out":0.22,"cacheRead":0.22,"uptime":100,"tps":106,"ttft":209,"reqs":10013}],"scores":{"bfcl":{"v":0.2583,"as":"Llama-3.1-8B-Instruct (Prompt)","effort":"","version":"V4","agent":"BFCL V4 · prompt-based tool calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":711,"as":"meta-llama/llama-3.1-8b-instruct","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":5,"tps":115,"ttft":221,"via":"Groq","reqs":52514},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3-30b-a3b-instruct-2507","name":"Qwen: Qwen3 30B A3B Instruct 2507","org":"Qwen","slug":"qwen3-30b-a3b-instruct-2507","matchKey":"qwen330ba3binstruct2507","blurb":"Qwen3-30B-A3B-Instruct-2507 is a 30.5B-parameter mixture-of-experts language model from Qwen, with 3.3B active parameters per inference.","created":1753806965,"cutoff":"2025-06-30","hf":"Qwen/Qwen3-30B-A3B-Instruct-2507","ctx":262144,"maxOut":32000,"in":["text"],"out":["text"],"price":{"in":0.04815,"out":0.19305},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"StreamLake","ctx":128000,"in":0.04815,"out":0.19305,"uptime":98.7,"tps":25,"ttft":1338,"reqs":15058},{"name":"DekaLLM","maxOut":235929,"in":0.09,"out":0.3,"uptime":99.2,"tps":51,"ttft":455,"reqs":875},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.09,"out":0.3,"uptime":94.8,"tps":15,"ttft":1276,"reqs":7504},{"name":"Nebius","quant":"fp8","maxOut":235929,"in":0.1,"out":0.3,"uptime":93.9,"tps":27,"ttft":363,"reqs":11563},{"name":"Alibaba","ctx":131072,"maxOut":32768,"in":0.13,"out":0.52,"uptime":100,"noSchema":1,"tps":35,"ttft":345,"reqs":5300}],"scores":{"arenaChat":{"v":1382.343474230352,"ci":4.9,"rank":173,"votes":23222,"as":"qwen3-30b-a3b-instruct-2507","effort":""},"bfcl":{"v":0.4139,"as":"Qwen3-30B-A3B-Instruct-2507 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":100,"n":5,"tps":25,"ttft":1338,"tpsBest":51,"via":"StreamLake","reqs":15058},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3-next-80b-a3b-instruct","name":"Qwen: Qwen3 Next 80B A3B Instruct","org":"Qwen","slug":"qwen3-next-80b-a3b-instruct","matchKey":"qwen3next80ba3binstruct","blurb":"Qwen3-Next-80B-A3B-Instruct is an instruction-tuned chat model in the Qwen3-Next series optimized for fast, stable responses without “thinking” traces.","created":1757612213,"cutoff":"2025-09-30","hf":"Qwen/Qwen3-Next-80B-A3B-Instruct","ctx":262144,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.09,"out":1.1},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.09,"out":1.1,"uptime":99.8,"tps":44,"ttft":921,"reqs":9890},{"name":"Alibaba","ctx":131072,"maxOut":32768,"in":0.0975,"out":0.78,"uptime":100,"noSchema":1,"tps":51,"ttft":478,"reqs":8142},{"name":"Parasail","quant":"fp8","maxOut":235929,"in":0.1,"out":1.1,"cacheRead":0.07,"uptime":100,"tps":56,"ttft":463,"reqs":8328},{"name":"Google","region":"global","maxOut":235929,"in":0.15,"out":1.2,"uptime":100,"tps":65,"ttft":474,"reqs":3333},{"name":"Novita","quant":"bf16","ctx":131072,"maxOut":32768,"in":0.15,"out":1.5,"uptime":99.8,"noSchema":1,"tps":26,"ttft":1140,"reqs":139}],"scores":{"arenaChat":{"v":1399.3282576722804,"ci":4.8,"rank":150,"votes":22529,"as":"qwen3-next-80b-a3b-instruct","effort":""},"eqLongform":{"v":31.5,"as":"Qwen/Qwen3-Next-80B-A3B-Instruct","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"regions":["global"],"n":5,"tps":44,"ttft":921,"tpsBest":65,"via":"DeepInfra","reqs":9890},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3-vl-235b-a22b-instruct","name":"Qwen: Qwen3 VL 235B A22B Instruct","org":"Qwen","slug":"qwen3-vl-235b-a22b-instruct","matchKey":"qwen3vl235ba22binstruct","blurb":"Qwen3-VL-235B-A22B Instruct is an open-weight multimodal model that unifies strong text generation with visual understanding across images and video.","created":1758668687,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-VL-235B-A22B-Instruct","ctx":262144,"maxOut":32768,"in":["text","image"],"out":["text"],"price":{"in":0.21,"out":1.9},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.2,"out":0.88,"cacheRead":0.11,"uptime":95.2,"tps":11,"ttft":1934,"reqs":976},{"name":"Venice","quant":"fp8","ctx":128000,"maxOut":16384,"in":0.21,"out":1.9,"cacheRead":0.1,"uptime":98.2,"tps":15,"ttft":2141,"reqs":305},{"name":"Parasail","quant":"fp8","ctx":131072,"in":0.21,"out":1.9,"cacheRead":0.1,"uptime":99.5,"tps":25,"ttft":1006,"reqs":1569},{"name":"Alibaba","ctx":131072,"in":0.26,"out":1.04,"uptime":99.8,"tps":27,"ttft":697,"reqs":4153},{"name":"Novita","quant":"bf16","ctx":131072,"in":0.3,"out":1.5,"uptime":96.9,"tps":26,"ttft":1933,"reqs":2486}],"scores":{"arenaChat":{"v":1414.068430473983,"ci":6.6,"rank":131,"votes":11219,"as":"qwen3-vl-235b-a22b-instruct","effort":""},"arenaVision":{"v":1214.734950626138,"ci":7.1,"rank":70,"votes":12250,"as":"qwen3-vl-235b-a22b-instruct","effort":""}},"perf":{"uptime":99.8,"cache":0.11,"n":5,"tps":27,"ttft":697,"via":"Alibaba","reqs":4153},"open":true,"rankable":true,"scored":2},{"id":"google/gemini-2.5-flash-image","name":"Google: Nano Banana (Gemini 2.5 Flash Image)","org":"Google","slug":"gemini-2.5-flash-image","matchKey":"gemini25flashimage","blurb":"Gemini 2.5 Flash Image, a.k.a.","created":1759870431,"cutoff":"2025-01-31","hf":null,"ctx":32768,"maxOut":8192,"in":["image","text"],"out":["image","text"],"price":{"in":0.3,"out":2.5,"imageOut":0.00003,"imageIn":3e-7,"audioIn":1,"webSearch":0.014},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google AI Studio","maxOut":29491,"in":0.15,"out":1.25,"cacheRead":0.015,"cacheWrite":0.041667,"uptime":100},{"name":"Google AI Studio","maxOut":29491,"in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":99.9},{"name":"Google","region":"global","in":0.3,"out":2.5,"cacheRead":0.03,"cacheWrite":0.083333,"uptime":100,"ttft":10162,"reqs":554},{"name":"Google AI Studio","maxOut":29491,"in":0.54,"out":4.5,"cacheRead":0.054,"cacheWrite":0.15,"uptime":100,"ttft":8195,"reqs":69}],"scores":{"arenaImageGen":{"v":1149.8594466609231,"ci":2.5,"rank":32,"votes":845086,"as":"gemini-2.5-flash-image-preview (nano-banana)","effort":""},"arenaImageEdit":{"v":1293.4062330794452,"ci":2.2,"rank":27,"votes":11176305,"as":"gemini-2.5-flash-image-preview (nano-banana)","effort":""}},"perf":{"uptime":100,"cache":0.015,"regions":["global"],"n":4,"ttft":10162,"via":"Google","reqs":554},"open":false,"rankable":true,"scored":2},{"id":"deepseek/deepseek-chat-v3-0324","name":"DeepSeek: DeepSeek V3 0324","org":"DeepSeek","slug":"deepseek-chat-v3-0324","matchKey":"deepseekchatv30324","blurb":"DeepSeek V3, a 685B-parameter, mixture-of-experts model, is the latest iteration of the flagship chat model family from the DeepSeek team.","created":1742824755,"cutoff":"2024-07-31","hf":"deepseek-ai/DeepSeek-V3-0324","ctx":163840,"maxOut":147456,"in":["text"],"out":["text"],"price":{"in":0.25,"out":1},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","maxOut":16384,"in":0.24,"out":0.9,"cacheRead":0.135,"uptime":85.5,"tps":21,"ttft":842,"reqs":2752},{"name":"SiliconFlow","quant":"fp8","in":0.25,"out":1,"uptime":95.1,"tps":22,"ttft":1545,"reqs":3590},{"name":"GMICloud","quant":"fp8","in":0.29,"out":1.14,"cacheRead":0.11,"uptime":95.6,"noTools":1,"tps":22,"ttft":2864,"reqs":5628}],"scores":{"eqWriting":{"v":1470.2,"as":"deepseek-ai/DeepSeek-V3-0324","version":"Creative Writing v3"},"eqLongform":{"v":60.1,"as":"deepseek-ai/DeepSeek-V3-0324","version":"Longform Writing v1.11"}},"perf":{"uptime":95.6,"cache":0.135,"n":3,"tps":22,"ttft":2864,"via":"GMICloud","reqs":5628},"open":true,"rankable":true,"scored":2},{"id":"mistralai/mistral-medium-3.1","name":"Mistral: Mistral Medium 3.1","org":"Mistral","slug":"mistral-medium-3.1","matchKey":"mistralmedium31","blurb":"Mistral Medium 3.1 is an updated version of Mistral Medium 3, which is a high-performance enterprise-grade language model designed to deliver frontier-level capabilities at significantly reduced operational cost.","created":1755095639,"cutoff":"2025-06-30","hf":null,"ctx":131072,"maxOut":104857,"in":["text","image","file"],"out":["text"],"price":{"in":0.4,"out":2,"alt":{"batch":{"in":0.2,"out":1}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.4,"out":2,"cacheRead":0.04,"uptime":99.6},{"name":"Mistral","in":0.4,"out":2,"cacheRead":0.04,"uptime":99.6},{"name":"Mistral","region":"eu","in":0.44,"out":2.2,"cacheRead":0.044,"uptime":99.9,"tps":71,"ttft":410,"reqs":60}],"scores":{"eqWriting":{"v":1474,"as":"mistral-medium-3.1","version":"Creative Writing v3"},"eqLongform":{"v":46.2,"as":"mistral-medium-3.1","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"cache":0.04,"regions":["eu"],"n":3,"tps":80,"ttft":250,"via":"Mistral","reqs":7102},"open":false,"rankable":true,"scored":2},{"id":"qwen/qwen3-14b","name":"Qwen: Qwen3 14B","org":"Qwen","slug":"qwen3-14b","matchKey":"qwen314b","blurb":"Qwen3-14B is a dense 14.8B parameter causal language model from the Qwen3 series, designed for both complex reasoning and efficient dialogue.","created":1745876478,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-14B","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.12,"out":0.24},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"NextBit","quant":"int4","ctx":40960,"maxOut":36864,"in":0.1,"out":0.22,"uptime":98.8,"noTools":1,"tps":67,"ttft":1236,"reqs":649},{"name":"DeepInfra","quant":"fp8","ctx":40960,"in":0.12,"out":0.24,"uptime":99,"tps":50,"ttft":1034,"reqs":841},{"name":"Alibaba","maxOut":8192,"in":0.2275,"out":0.91,"uptime":100,"noSchema":1,"tps":63,"ttft":310,"reqs":195}],"scores":{"bfcl":{"v":0.4103,"as":"Qwen3-14B (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqLongform":{"v":35.9,"as":"qwen/qwen3-14b","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":3,"tps":50,"ttft":1034,"tpsBest":67,"via":"DeepInfra","reqs":841},"open":true,"rankable":true,"scored":2},{"id":"amazon/nova-2-lite-v1","name":"Amazon: Nova 2 Lite","org":"Amazon","slug":"nova-2-lite-v1","matchKey":"nova2litev1","blurb":"Nova 2 Lite is a fast, cost-effective reasoning model for everyday workloads that can process text, images, and videos to generate text.","created":1764696672,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65535,"in":["text","image","video","file"],"out":["text"],"price":{"in":0.3,"out":2.5},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"Amazon Bedrock","region":"eu","in":0.3,"out":2.5,"uptime":98.9},{"name":"Amazon Bedrock","in":0.3,"out":2.5,"uptime":99,"tps":109,"ttft":1178,"reqs":612,"region":"eu"}],"scores":{"bfcl":{"v":0.271,"as":"Amazon-Nova-2-Lite-v1:0 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqLongform":{"v":38.8,"as":"nova-2-lite-v1","version":"Longform Writing v1.11"}},"perf":{"uptime":99,"regions":["eu"],"n":2,"tps":109,"ttft":1178,"via":"Amazon Bedrock","reqs":612},"open":false,"rankable":true,"scored":2},{"id":"google/gemini-3-pro-image","name":"Google: Nano Banana Pro (Gemini 3 Pro Image)","org":"Google","slug":"gemini-3-pro-image","matchKey":"gemini3proimage","blurb":"Nano Banana Pro is Google’s most advanced image-generation and editing model, built on Gemini 3 Pro.","created":1781754054,"cutoff":null,"hf":null,"ctx":131072,"maxOut":32768,"in":["image","text"],"out":["image","text"],"price":{"in":2,"out":12,"imageOut":0.00012,"imageIn":0.000002,"audioIn":2,"webSearch":0.014},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Google AI Studio","region":"global","in":2,"out":12,"cacheRead":0.2,"cacheWrite":0.375,"uptime":100,"ttft":24832,"reqs":78},{"name":"Google","region":"global","ctx":65536,"in":2,"out":12,"cacheRead":0.2,"cacheWrite":0.375,"uptime":99.6,"noTools":1,"ttft":26832,"reqs":366}],"scores":{"arenaImageGen":{"v":1246.241435218239,"ci":3.2,"rank":14,"votes":152220,"as":"gemini-3-pro-image-2k (nano-banana-pro)","effort":""},"arenaImageEdit":{"v":1390.0365753708506,"ci":2.7,"rank":9,"votes":556580,"as":"gemini-3-pro-image-2k (nano-banana-pro)","effort":""}},"perf":{"uptime":100,"cache":0.2,"regions":["global"],"n":2,"ttft":26832,"via":"Google","reqs":366},"open":false,"rankable":true,"scored":2},{"id":"google/gemini-3-pro-image-preview","name":"Google: Nano Banana Pro (Gemini 3 Pro Image Preview)","org":"Google","slug":"gemini-3-pro-image-preview","matchKey":"gemini3proimagepreview","blurb":"Nano Banana Pro is Google’s most advanced image-generation and editing model, built on Gemini 3 Pro.","created":1763653797,"cutoff":null,"hf":null,"ctx":65536,"maxOut":32768,"in":["image","text"],"out":["image","text"],"price":{"in":2,"out":12,"imageOut":0.00012,"imageIn":0.000002,"audioIn":2,"webSearch":0.014},"caps":{"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Google AI Studio","in":1,"out":6,"cacheRead":0.1,"cacheWrite":0.1875,"uptime":99.9},{"name":"Google AI Studio","region":"global","in":2,"out":12,"cacheRead":0.2,"cacheWrite":0.375,"uptime":100,"ttft":23462,"reqs":432}],"scores":{"arenaImageGen":{"v":1232.1582152667036,"ci":4.9,"rank":16,"votes":82737,"as":"gemini-3-pro-image-preview (nano-banana-pro)","effort":""},"arenaImageEdit":{"v":1385.451415510249,"ci":2.9,"rank":13,"votes":518495,"as":"gemini-3-pro-image-preview (nano-banana-pro)","effort":""}},"perf":{"uptime":100,"cache":0.1,"regions":["global"],"n":2,"ttft":23462,"via":"Google AI Studio","reqs":432},"open":false,"rankable":true,"scored":2},{"id":"google/gemini-3.1-flash-image","name":"Google: Nano Banana 2 (Gemini 3.1 Flash Image)","org":"Google","slug":"gemini-3.1-flash-image","matchKey":"gemini31flashimage","blurb":"Gemini 3.1 Flash Image, a.k.a.","created":1781754065,"cutoff":null,"hf":null,"ctx":131072,"maxOut":32768,"in":["image","text"],"out":["image","text"],"price":{"in":0.5,"out":3,"imageOut":0.00006,"webSearch":0.014},"caps":{"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","minimal"]},"moderated":false,"providers":[{"name":"Google AI Studio","ctx":65536,"maxOut":58982,"in":0.5,"out":3,"uptime":99.9,"ttft":16703,"reqs":95},{"name":"Google","region":"global","in":0.5,"out":3,"uptime":99.9,"noSchema":1,"ttft":18492,"reqs":3014}],"scores":{"arenaImageGen":{"v":1260.680952945605,"ci":4.6,"rank":9,"votes":41957,"as":"gemini-3.1-flash-image (nano-banana-2) [web-search]","effort":""},"arenaImageEdit":{"v":1387.492132534014,"ci":3.5,"rank":12,"votes":157693,"as":"gemini-3.1-flash-image (nano-banana-2) [web-search]","effort":""}},"perf":{"uptime":99.9,"regions":["global"],"n":2,"ttft":18492,"via":"Google","reqs":3014},"open":false,"rankable":true,"scored":2},{"id":"google/gemini-3.1-flash-lite-image","name":"Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)","org":"Google","slug":"gemini-3.1-flash-lite-image","matchKey":"gemini31flashliteimage","blurb":"Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is Google's fastest, most cost-efficient Gemini image model, built for high-velocity developer pipelines and rapid-fire visual exploration.","created":1782837225,"cutoff":"2025-01-01","hf":null,"ctx":65536,"maxOut":58982,"in":["image","text"],"out":["image","text"],"price":{"in":0.25,"out":1.5,"imageOut":0.00003,"webSearch":0.014},"caps":{"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","minimal"]},"moderated":false,"providers":[{"name":"Google","region":"global","in":0.25,"out":1.5,"uptime":100,"ttft":7693,"reqs":2151},{"name":"Google AI Studio","in":0.25,"out":1.5,"uptime":100,"ttft":5644,"reqs":78}],"scores":{"arenaImageGen":{"v":1250.402349022443,"ci":6.2,"rank":13,"votes":16371,"as":"gemini-3.1-flash-lite-image (nano-banana-2-lite)","effort":""},"arenaImageEdit":{"v":1314.295914743134,"ci":4.9,"rank":19,"votes":59280,"as":"gemini-3.1-flash-lite-image (nano-banana-2-lite)","effort":""}},"perf":{"uptime":100,"regions":["global"],"n":2,"ttft":7693,"via":"Google","reqs":2151},"open":false,"rankable":true,"scored":2},{"id":"meta-llama/llama-3.2-3b-instruct","name":"Meta: Llama 3.2 3B Instruct","org":"Meta","slug":"llama-3.2-3b-instruct","matchKey":"llama323binstruct","blurb":"Llama 3.2 3B is a 3-billion-parameter multilingual large language model, optimized for advanced natural language processing tasks like dialogue generation, reasoning, and summarization.","created":1727222400,"cutoff":"2023-12-31","hf":"meta-llama/Llama-3.2-3B-Instruct","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.05,"out":0.33},"caps":{"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"bf16","in":0.05,"out":0.33,"uptime":99.9,"tps":58,"ttft":480,"reqs":1895},{"name":"Cloudflare","ctx":80000,"maxOut":72000,"in":0.0509,"out":0.335,"uptime":99.5,"noSchema":1,"tps":54,"ttft":198,"reqs":1310}],"scores":{"bfcl":{"v":0.2195,"as":"Llama-3.2-3B-Instruct (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":593.5,"as":"meta-llama/llama-3.2-3b-instruct","version":"Creative Writing v3"}},"perf":{"uptime":99.9,"n":2,"tps":58,"ttft":480,"via":"Parasail","reqs":1895},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3-32b","name":"Qwen: Qwen3 32B","org":"Qwen","slug":"qwen3-32b","matchKey":"qwen332b","blurb":"Qwen3-32B is a dense 32.8B parameter causal language model from the Qwen3 series, optimized for both complex reasoning and efficient dialogue.","created":1745875945,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-32B","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.08,"out":0.28},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","ctx":40960,"in":0.08,"out":0.28,"uptime":99.9,"tps":25,"ttft":339,"reqs":22518},{"name":"SiliconFlow","quant":"fp8","maxOut":117964,"in":0.14,"out":0.57,"uptime":98.2,"tps":29,"ttft":1017,"reqs":978}],"scores":{"bfcl":{"v":0.48710000000000003,"as":"Qwen3-32B (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqLongform":{"v":40.5,"as":"qwen/qwen3-32b","version":"Longform Writing v1.11"}},"perf":{"uptime":99.9,"n":2,"tps":25,"ttft":339,"tpsBest":29,"via":"DeepInfra","reqs":22518},"open":true,"rankable":true,"scored":2},{"id":"qwen/qwen3-vl-235b-a22b-thinking","name":"Qwen: Qwen3 VL 235B A22B Thinking","org":"Qwen","slug":"qwen3-vl-235b-a22b-thinking","matchKey":"qwen3vl235ba22bthinking","blurb":"Qwen3-VL-235B-A22B Thinking is a multimodal model that unifies strong text generation with visual understanding across images and video.","created":1758668690,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-VL-235B-A22B-Thinking","ctx":131072,"maxOut":32768,"in":["text","image"],"out":["text"],"price":{"in":0.4,"out":4},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.4,"out":4,"uptime":93.4,"tps":55,"ttft":559,"reqs":71},{"name":"Novita","quant":"bf16","in":0.98,"out":3.95,"uptime":99.8,"noSchema":1,"tps":47,"ttft":958,"reqs":19}],"scores":{"arenaChat":{"v":1395.1502102479446,"ci":6.9,"rank":154,"votes":7780,"as":"qwen3-vl-235b-a22b-thinking","effort":"default"},"arenaVision":{"v":1187.553683920701,"ci":12.4,"rank":87,"votes":2389,"as":"qwen3-vl-235b-a22b-thinking","effort":"default"}},"perf":{"uptime":99.8,"n":2,"tps":55,"ttft":559,"via":"Alibaba","reqs":71},"open":true,"rankable":true,"scored":2},{"id":"upstage/solar-pro4","name":"Upstage: Solar Pro 4","org":"Upstage","slug":"solar-pro4","matchKey":"solarpro4","blurb":"Solar Pro 4 is Upstage's cost-efficient large language model, featuring a 524K context window.","created":1786371636,"cutoff":null,"hf":null,"ctx":524288,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.09,"out":0.36},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","minimal","none"]},"moderated":false,"providers":[{"name":"Upstage","in":0.09,"out":0.36,"cacheRead":0.018,"uptime":99.2},{"name":"Upstage","in":0.09,"out":0.36,"cacheRead":0.018,"uptime":99.2,"tps":47,"ttft":2149,"reqs":50188}],"scores":{"arenaChat":{"v":1376.4070978522318,"ci":12.2,"rank":177,"votes":2420,"as":"solar-pro4","effort":""},"arenaWebdev":{"v":1371.4775097721108,"ci":16.9,"rank":88,"votes":1403,"as":"solar-pro4","effort":""}},"perf":{"uptime":99.2,"cache":0.018,"n":2,"tps":47,"ttft":2149,"via":"Upstage","reqs":50188},"open":false,"rankable":true,"scored":2},{"id":"z-ai/glm-4.5v","name":"Z.ai: GLM 4.5V","org":"Z.ai","slug":"glm-4.5v","matchKey":"glm45v","blurb":"GLM-4.5V is a vision-language foundation model for multimodal agent applications.","created":1754922288,"cutoff":"2024-12-31","hf":"zai-org/GLM-4.5V","ctx":65536,"maxOut":16384,"in":["text","image"],"out":["text"],"price":{"in":0.6,"out":1.8},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"fp8","in":0.6,"out":1.8,"cacheRead":0.11,"uptime":95.1,"tps":18,"ttft":1729,"reqs":45},{"name":"Z.AI","quant":"fp8","in":0.6,"out":1.8,"cacheRead":0.11,"uptime":97.7,"tps":30,"ttft":3170,"reqs":101}],"scores":{"arenaChat":{"v":1352.5539384618924,"ci":8.4,"rank":200,"votes":4817,"as":"glm-4.5v","effort":""},"arenaVision":{"v":1153.539697508655,"ci":11.8,"rank":103,"votes":3255,"as":"glm-4.5v","effort":""}},"perf":{"uptime":97.7,"cache":0.11,"n":2,"tps":30,"ttft":3170,"via":"Z.AI","reqs":101},"open":true,"rankable":true,"scored":2},{"id":"z-ai/glm-4.6v","name":"Z.ai: GLM 4.6V","org":"Z.ai","slug":"glm-4.6v","matchKey":"glm46v","blurb":"GLM-4.6V is a large multimodal model designed for high-fidelity visual understanding and long-context reasoning across images, documents, and mixed media.","created":1765207462,"cutoff":null,"hf":"zai-org/GLM-4.6V","ctx":131072,"maxOut":32768,"in":["image","text","video"],"out":["text"],"price":{"in":0.3,"out":0.9},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"bf16","in":0.3,"out":0.9,"cacheRead":0.055,"uptime":99.2,"tps":15,"ttft":1907,"reqs":306},{"name":"Z.AI","quant":"fp8","in":0.3,"out":0.9,"cacheRead":0.05,"uptime":98.7,"tps":17,"ttft":5854,"reqs":312}],"scores":{"arenaChat":{"v":1378.4585966065274,"ci":11.3,"rank":175,"votes":2762,"as":"glm-4.6v","effort":""},"arenaVision":{"v":1160.9688922280434,"ci":14,"rank":97,"votes":2352,"as":"glm-4.6v","effort":""}},"perf":{"uptime":99.2,"cache":0.055,"n":2,"tps":17,"ttft":5854,"via":"Z.AI","reqs":312},"open":true,"rankable":true,"scored":2},{"id":"alibaba/happyhorse-1.0","name":"Alibaba: HappyHorse 1.0","org":"Alibaba","slug":"happyhorse-1.0","matchKey":"happyhorse10","blurb":"HappyHorse 1.0 is a video generation model from Alibaba.","created":1782260324,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0,"out":0}],"scores":{"arenaVideoGen":{"v":1427.403188202753,"ci":12.5,"rank":10,"votes":22116,"as":"happyhorse-1.0","effort":""},"arenaVideoEdit":{"v":1441.602760829705,"ci":9.8,"rank":10,"votes":70658,"as":"happyhorse-1.0","effort":""}},"perf":{"n":1},"open":false,"rankable":true,"scored":2},{"id":"alibaba/wan-3.0","name":"Alibaba: Wan 3.0","org":"Alibaba","slug":"wan-3.0","matchKey":"wan30","blurb":"Wan 3.0 is a video generation model from Alibaba for text-to-video, image-to-video, and reference-guided video generation.","created":1787603856,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0,"out":0,"uptime":100,"ttft":155540,"reqs":8}],"scores":{"arenaVideoGen":{"v":1494.3428809079605,"ci":19.4,"rank":3,"votes":1167,"as":"wan3.0","effort":""},"arenaVideoEdit":{"v":1479.4157079063789,"ci":10.8,"rank":3,"votes":3444,"as":"wan3.0","effort":""}},"perf":{"uptime":100,"n":1,"ttft":155540,"via":"Alibaba","reqs":8},"open":false,"rankable":true,"scored":2},{"id":"anthropic/claude-3-haiku","name":"Anthropic: Claude 3 Haiku","org":"Anthropic","slug":"claude-3-haiku","matchKey":"claude3haiku","blurb":"Claude 3 Haiku is Anthropic's fastest and most compact model for\nnear-instant responsiveness.","created":1710288000,"cutoff":"2023-08-31","hf":null,"ctx":200000,"maxOut":4096,"in":["text","image"],"out":["text"],"price":{"in":0.25,"out":1.25,"webSearch":0.01},"caps":{"tools":true},"reasoning":null,"moderated":true,"providers":[{"name":"Amazon Bedrock","in":0.25,"out":1.25,"cacheRead":0.03,"cacheWrite":0.3,"uptime":100,"tps":68,"ttft":346,"reqs":2733}],"scores":{"arenaVision":{"v":1000.4884218850431,"ci":12.4,"rank":141,"votes":13380,"as":"claude-3-haiku-20240307","effort":""},"eqWriting":{"v":714.6,"as":"anthropic/claude-3-haiku","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.03,"n":1,"tps":68,"ttft":346,"via":"Amazon Bedrock","reqs":2733},"open":false,"rankable":true,"scored":2},{"id":"arcee-ai/trinity-large-thinking","name":"Arcee AI: Trinity Large Thinking","org":"Arcee AI","slug":"trinity-large-thinking","matchKey":"trinitylargethinking","blurb":"Trinity Large Thinking is a powerful open source reasoning model from the team at Arcee AI.","created":1775058318,"cutoff":null,"hf":"arcee-ai/Trinity-Large-Thinking","ctx":262144,"maxOut":80000,"in":["text"],"out":["text"],"price":{"in":0.25,"out":0.8},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Arcee AI","in":0.25,"out":0.8,"cacheRead":0.06,"uptime":100,"tps":80,"ttft":165,"reqs":481}],"scores":{"arenaChat":{"v":1378.4441472224296,"ci":4.4,"rank":176,"votes":29742,"as":"trinity-large-preview","effort":""},"arenaWebdev":{"v":1236.6066793515347,"ci":20.5,"rank":119,"votes":1303,"as":"trinity-large-thinking","effort":"default"}},"perf":{"uptime":100,"cache":0.06,"n":1,"tps":80,"ttft":165,"via":"Arcee AI","reqs":481},"open":true,"rankable":true,"scored":2},{"id":"black-forest-labs/flux-3-video","name":"Black Forest Labs: FLUX.3 Video","org":"Black Forest Labs","slug":"flux-3-video","matchKey":"flux3video","blurb":"FLUX.3 Video is a video generation model from Black Forest Labs.","created":1785858831,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","in":0,"out":0,"uptime":100,"ttft":302680,"reqs":2}],"scores":{"arenaVideoGen":{"v":1493.8040842194478,"ci":17.3,"rank":4,"votes":1290,"as":"flux-3-video","effort":""},"arenaVideoEdit":{"v":1449.9843486836905,"ci":5.4,"rank":9,"votes":30684,"as":"flux-3-video-20260811","effort":""}},"perf":{"uptime":100,"n":1,"ttft":302680,"via":"Black Forest Labs","reqs":2},"open":false,"rankable":true,"scored":2},{"id":"black-forest-labs/flux.2-flex","name":"Black Forest Labs: FLUX.2 Flex","org":"Black Forest Labs","slug":"flux.2-flex","matchKey":"flux2flex","blurb":"FLUX.2 [flex] excels at rendering complex text, typography, and fine details, and supports multi-reference editing in the same unified architecture.","created":1764045987,"cutoff":null,"hf":null,"ctx":67344,"maxOut":60609,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000146484375},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","region":"us","in":0,"out":0,"uptime":100}],"scores":{"arenaImageGen":{"v":1156.5030849484722,"ci":3.4,"rank":27,"votes":149288,"as":"flux-2-flex","effort":""},"arenaImageEdit":{"v":1225.2408504002967,"ci":3.2,"rank":38,"votes":415750,"as":"flux-2-flex","effort":""}},"perf":{"uptime":100,"regions":["us"],"n":1},"open":false,"rankable":true,"scored":2},{"id":"black-forest-labs/flux.2-klein-4b","name":"Black Forest Labs: FLUX.2 Klein 4B","org":"Black Forest Labs","slug":"flux.2-klein-4b","matchKey":"flux2klein4b","blurb":"FLUX.2 [klein] 4B is the fastest and most cost-effective model in the FLUX.2 family, optimized for high-throughput use cases while maintaining excellent image quality.","created":1768429228,"cutoff":null,"hf":"black-forest-labs/FLUX.2-klein-4B","ctx":40960,"maxOut":36864,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000341796875},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","in":0,"out":0,"uptime":100,"ttft":3673,"reqs":117}],"scores":{"arenaImageGen":{"v":1029.9861799619957,"ci":3.3,"rank":66,"votes":144483,"as":"flux-2-klein-4b","effort":""},"arenaImageEdit":{"v":1187.7794144112104,"ci":3,"rank":45,"votes":507119,"as":"flux-2-klein-4b","effort":""}},"perf":{"uptime":100,"n":1,"ttft":3673,"via":"Black Forest Labs","reqs":117},"open":true,"rankable":true,"scored":2},{"id":"black-forest-labs/flux.2-max","name":"Black Forest Labs: FLUX.2 Max","org":"Black Forest Labs","slug":"flux.2-max","matchKey":"flux2max","blurb":"FLUX.2 [max] is the new top-tier image model from Black Forest Labs, pushing image quality, prompt understanding, and editing consistency to the highest level yet.","created":1765857570,"cutoff":null,"hf":null,"ctx":46864,"maxOut":42177,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00001708984375},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","region":"us","in":0,"out":0,"uptime":100,"ttft":48739,"reqs":9}],"scores":{"arenaImageGen":{"v":1162.0418393816763,"ci":3.6,"rank":25,"votes":117339,"as":"flux-2-max","effort":"max"},"arenaImageEdit":{"v":1261.8104616919827,"ci":3.2,"rank":29,"votes":352883,"as":"flux-2-max","effort":"max"}},"perf":{"uptime":100,"regions":["us"],"n":1,"ttft":48739,"via":"Black Forest Labs","reqs":9},"open":false,"rankable":true,"scored":2},{"id":"black-forest-labs/flux.2-pro","name":"Black Forest Labs: FLUX.2 Pro","org":"Black Forest Labs","slug":"flux.2-pro","matchKey":"flux2pro","blurb":"A high-end image generation and editing model focused on frontier-level visual quality and reliability.","created":1764030274,"cutoff":null,"hf":null,"ctx":46864,"maxOut":42177,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000732421875},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","in":0,"out":0,"uptime":100,"ttft":12364,"reqs":130}],"scores":{"arenaImageGen":{"v":1153.9261650549058,"ci":3,"rank":28,"votes":188083,"as":"flux-2-pro","effort":""},"arenaImageEdit":{"v":1244.700228980646,"ci":2.7,"rank":32,"votes":539449,"as":"flux-2-pro","effort":""}},"perf":{"uptime":100,"n":1,"ttft":12364,"via":"Black Forest Labs","reqs":130},"open":false,"rankable":true,"scored":2},{"id":"bytedance-seed/seedream-4.5","name":"ByteDance Seed: Seedream 4.5","org":"ByteDance Seed","slug":"seedream-4.5","matchKey":"seedream45","blurb":"Seedream 4.5 is the latest in-house image generation model developed by ByteDance.","created":1766519506,"cutoff":null,"hf":null,"ctx":4096,"maxOut":3686,"in":["image","text"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000958083832335329},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100,"ttft":11843,"reqs":210}],"scores":{"arenaImageGen":{"v":1147.3450792972574,"ci":2.7,"rank":34,"votes":259145,"as":"seedream-4.5","effort":""},"arenaImageEdit":{"v":1301.5114450285334,"ci":2.3,"rank":24,"votes":1100289,"as":"seedream-4.5","effort":""}},"perf":{"uptime":100,"n":1,"ttft":11843,"via":"Seed","reqs":210},"open":false,"rankable":true,"scored":2},{"id":"bytedance-seed/seedream-5-0-lite","name":"ByteDance Seed: Seedream 5.0 Lite","org":"ByteDance Seed","slug":"seedream-5-0-lite","matchKey":"seedream50lite","blurb":"Seedream 5.0 Lite is an image generation model from ByteDance Seed.","created":1786650094,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000838323353293413},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100,"ttft":33213,"reqs":65}],"scores":{"arenaImageGen":{"v":1137.668893672582,"ci":3.4,"rank":37,"votes":109182,"as":"seedream-5.0-lite","effort":""},"arenaImageEdit":{"v":1293.8360594492444,"ci":2.9,"rank":26,"votes":373781,"as":"seedream-5.0-lite","effort":""}},"perf":{"uptime":100,"n":1,"ttft":33213,"via":"Seed","reqs":65},"open":false,"rankable":true,"scored":2},{"id":"bytedance-seed/seedream-5-0-pro","name":"ByteDance Seed: Seedream 5.0 Pro","org":"ByteDance Seed","slug":"seedream-5-0-pro","matchKey":"seedream50pro","blurb":"Seedream 5.0 Pro is an image generation and editing model from ByteDance Seed.","created":1786578139,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000107784431137725,"imageIn":0.003},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100,"ttft":61985,"reqs":89}],"scores":{"arenaImageGen":{"v":1257.1038107318113,"ci":4.3,"rank":10,"votes":62443,"as":"seedream-5.0-pro","effort":""},"arenaImageEdit":{"v":1394.483446178541,"ci":3.6,"rank":8,"votes":179966,"as":"seedream-5.0-pro","effort":""}},"perf":{"uptime":100,"n":1,"ttft":61985,"via":"Seed","reqs":89},"open":false,"rankable":true,"scored":2},{"id":"cohere/command-a","name":"Cohere: Command A","org":"Cohere","slug":"command-a","matchKey":"commanda","blurb":"Command A is an open-weights 111B parameter model with a 256k context window focused on delivering great performance across agentic, multilingual, and coding use cases.","created":1741894342,"cutoff":"2024-08-31","hf":"CohereForAI/c4ai-command-a-03-2025","ctx":256000,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":2.5,"out":10},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Cohere","in":2.5,"out":10,"uptime":100,"tps":58,"ttft":245,"reqs":81}],"scores":{"bfcl":{"v":0.46490000000000004,"as":"Command A (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":1142.8,"as":"CohereForAI/c4ai-command-a-03-2025","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":1,"tps":58,"ttft":245,"via":"Cohere","reqs":81},"open":true,"rankable":true,"scored":2},{"id":"meta-llama/llama-3.2-1b-instruct","name":"Meta: Llama 3.2 1B Instruct","org":"Meta","slug":"llama-3.2-1b-instruct","matchKey":"llama321binstruct","blurb":"Llama 3.2 1B is a 1-billion-parameter language model focused on efficiently performing natural language tasks, such as summarization, dialogue, and multilingual text analysis.","created":1727222400,"cutoff":"2023-12-31","hf":"meta-llama/Llama-3.2-1B-Instruct","ctx":60000,"maxOut":54000,"in":["text"],"out":["text"],"price":{"in":0.027,"out":0.201},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Cloudflare","in":0.027,"out":0.201,"uptime":100,"tps":22,"ttft":200,"reqs":76}],"scores":{"bfcl":{"v":0.1082,"as":"Llama-3.2-1B-Instruct (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"},"eqWriting":{"v":200,"as":"meta-llama/llama-3.2-1b-instruct","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":1,"tps":22,"ttft":200,"via":"Cloudflare","reqs":76},"open":true,"rankable":true,"scored":2},{"id":"meta/muse-image","name":"Meta: Muse Image","org":"Meta","slug":"muse-image","matchKey":"museimage","blurb":"Muse Image is an agentic image generation model from Meta that generates and edits images from text and reference images.","created":1787764532,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000239520958083832},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"Meta","in":0,"out":0,"uptime":99.8,"ttft":20084,"reqs":331}],"scores":{"arenaImageGen":{"v":1277.3049884141953,"ci":5.9,"rank":7,"votes":26045,"as":"muse-image","effort":""},"arenaImageEdit":{"v":1403.227284484497,"ci":5.1,"rank":6,"votes":83408,"as":"muse-image","effort":""}},"perf":{"uptime":99.8,"n":1,"ttft":20084,"via":"Meta","reqs":331},"open":false,"rankable":true,"scored":2},{"id":"microsoft/mai-image-2.5","name":"Microsoft AI: MAI-Image-2.5","org":"Microsoft AI","slug":"mai-image-2.5","matchKey":"maiimage25","blurb":"Microsoft AI's MAI-Image-2.5 is a high-quality image generation model available via Azure AI Foundry.","created":1780424896,"cutoff":null,"hf":null,"ctx":4096,"maxOut":1024,"in":["text","image"],"out":["image"],"price":{"in":5,"out":0,"imageOut":0.000047},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":5,"out":0,"uptime":100,"ttft":23271,"reqs":19}],"scores":{"arenaImageGen":{"v":1254.0716694351397,"ci":4.3,"rank":12,"votes":57250,"as":"mai-image-2.5","effort":""},"arenaImageEdit":{"v":1400.2099425389276,"ci":3.8,"rank":7,"votes":174744,"as":"mai-image-2.5","effort":""}},"perf":{"uptime":100,"n":1,"ttft":23271,"via":"Azure","reqs":19},"open":false,"rankable":true,"scored":2},{"id":"microsoft/mai-image-2.6","name":"Microsoft AI: MAI-Image-2.6","org":"Microsoft AI","slug":"mai-image-2.6","matchKey":"maiimage26","blurb":"MAI-Image-2.6 is an image generation and editing model from Microsoft AI, the precision tier of the MAI-Image-2.6 family alongside the faster [MAI-Image-2.6 Flash](/microsoft/mai-image-2.6-flash).","created":1788550742,"cutoff":null,"hf":null,"ctx":4096,"maxOut":1024,"in":["text","image"],"out":["image"],"price":{"in":5,"out":0,"imageOut":0.000038},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":5,"out":0,"uptime":100,"ttft":22572,"reqs":33}],"scores":{"arenaImageGen":{"v":1330.7442148973387,"ci":6.9,"rank":4,"votes":11213,"as":"mai-image-2.6","effort":""},"arenaImageEdit":{"v":1433.916630671604,"ci":7.6,"rank":5,"votes":9481,"as":"mai-image-2.6","effort":""}},"perf":{"uptime":100,"n":1,"ttft":22572,"via":"Azure","reqs":33},"open":false,"rankable":true,"scored":2},{"id":"minimax/hailuo-2.3","name":"MiniMax: Hailuo 2.3","org":"MiniMax","slug":"hailuo-2.3","matchKey":"hailuo23","blurb":"Hailuo 2.3 is a video generation model from MiniMax.","created":1776702740,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Minimax","in":0,"out":0,"ttft":203531,"reqs":1}],"scores":{"arenaVideoGen":{"v":1205.332677108679,"ci":6.1,"rank":32,"votes":82971,"as":"hailuo-2.3","effort":""},"arenaVideoEdit":{"v":1261.7825198302162,"ci":4.9,"rank":29,"votes":394450,"as":"hailuo-2.3","effort":""}},"perf":{"n":1,"ttft":203531,"via":"Minimax","reqs":1},"open":false,"rankable":true,"scored":2},{"id":"mistralai/devstral-2512","name":"Mistral: Devstral 2 2512","org":"Mistral","slug":"devstral-2512","matchKey":"devstral2512","blurb":"Devstral 2 is a state-of-the-art open-source model by Mistral AI specializing in agentic coding.","created":1765285419,"cutoff":null,"hf":"mistralai/Devstral-2-123B-Instruct-2512","ctx":262144,"maxOut":209715,"in":["text","file"],"out":["text"],"price":{"in":0.4,"out":2},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","region":"eu","in":0.4,"out":2,"cacheRead":0.04,"uptime":100,"tps":134,"ttft":384,"reqs":3868}],"scores":{"arenaWebdev":{"v":1081.1775273979379,"ci":30.9,"rank":128,"votes":831,"as":"devstral-medium-2507","effort":"medium"},"swebenchAgent":{"v":0.5379999999999999,"as":"Devstral (2512)","agent":"mini-SWE-agent","date":"2025-12-09","checked":true}},"perf":{"uptime":100,"cache":0.04,"regions":["eu"],"n":1,"tps":134,"ttft":384,"via":"Mistral","reqs":3868},"open":true,"rankable":true,"scored":2},{"id":"mistralai/mistral-small-3.1-24b-instruct","name":"Mistral: Mistral Small 3.1 24B","org":"Mistral","slug":"mistral-small-3.1-24b-instruct","matchKey":"mistralsmall3124binstruct","blurb":"Mistral Small 3.1 24B Instruct is an upgraded variant of Mistral Small 3 (2501), featuring 24 billion parameters with advanced multimodal capabilities.","created":1742238937,"cutoff":"2023-10-31","hf":"mistralai/Mistral-Small-3.1-24B-Instruct-2503","ctx":128000,"maxOut":102400,"in":["text","image"],"out":["text"],"price":{"in":0.351,"out":0.555},"caps":{"tools":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Cloudflare","in":0.351,"out":0.555,"uptime":100,"tps":15,"ttft":289,"reqs":1261}],"scores":{"arenaVision":{"v":1127.45217517799,"ci":8.5,"rank":112,"votes":28542,"as":"mistral-small-3.1-24b-instruct-2503","effort":""},"eqWriting":{"v":759.2,"as":"mistralai/mistral-small-3.1-24b-instruct-2503","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":1,"tps":15,"ttft":289,"via":"Cloudflare","reqs":1261},"open":true,"rankable":true,"scored":2},{"id":"moonshotai/kimi-k2-0905","name":"MoonshotAI: Kimi K2 0905","org":"MoonshotAI","slug":"kimi-k2-0905","matchKey":"kimik20905","blurb":"Kimi K2 0905 is the September update of [Kimi K2 0711](moonshotai/kimi-k2).","created":1757021147,"cutoff":"2024-12-31","hf":"moonshotai/Kimi-K2-Instruct-0905","ctx":262144,"maxOut":98304,"in":["text"],"out":["text"],"price":{"in":0.6,"out":2.5},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Novita","quant":"fp8","in":0.6,"out":2.5,"uptime":100,"tps":35,"ttft":660,"reqs":2115}],"scores":{"arenaChat":{"v":1418.055969275088,"ci":6.5,"rank":120,"votes":11568,"as":"kimi-k2-0905-preview","effort":""},"eqLongform":{"v":63.8,"as":"moonshotai/Kimi-K2-Instruct-0905","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":1,"tps":35,"ttft":660,"via":"Novita","reqs":2115},"open":true,"prose":{"slop":10.118758416584026,"trigram":0.07764888063592215,"notXButY":0.1743028156360385,"repetition":3.9579079629338776,"mattr":0.6237694603863136,"grade":9.436166980762671,"sentence":17.24724907063197,"words":450747,"samples":300,"as":"moonshotai/kimi-k2-0905"},"rankable":true,"scored":2},{"id":"nousresearch/hermes-4-405b","name":"Nous: Hermes 4 405B","org":"Nous","slug":"hermes-4-405b","matchKey":"hermes4405b","blurb":"Hermes 4 is a large-scale reasoning model built on Meta-Llama-3.1-405B and released by Nous Research.","created":1756235463,"cutoff":"2024-08-31","hf":"NousResearch/Hermes-4-405B","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":1,"out":3},"caps":{"json":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Nebius","quant":"fp8","in":1,"out":3,"uptime":100,"tps":36,"ttft":248,"reqs":306}],"scores":{"eqWriting":{"v":1418.5,"as":"NousResearch/Hermes-4-405B","version":"Creative Writing v3"},"eqLongform":{"v":51.2,"as":"NousResearch/Hermes-4-405B","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":1,"tps":36,"ttft":248,"via":"Nebius","reqs":306},"open":true,"rankable":true,"scored":2},{"id":"openai/gpt-4-turbo","name":"OpenAI: GPT-4 Turbo","org":"OpenAI","slug":"gpt-4-turbo","matchKey":"gpt4turbo","blurb":"The latest GPT-4 Turbo model with vision capabilities.","created":1712620800,"cutoff":"2023-12-31","hf":null,"ctx":128000,"maxOut":4096,"in":["text","image"],"out":["text"],"price":{"in":10,"out":30,"alt":{"batch":{"in":5,"out":15}}},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":10,"out":30,"uptime":100,"tps":4,"ttft":1158,"reqs":18}],"scores":{"arenaVision":{"v":1112.1668804673986,"ci":11.6,"rank":120,"votes":13391,"as":"gpt-4-turbo-2024-04-09","effort":""},"gpqa":{"v":0.4236111111111111,"as":"GPT-4 Turbo Preview (Nov 2023)","effort":"","date":"2025-01-27","stderr":0.024076142618301676}},"perf":{"uptime":100,"n":1,"tps":4,"ttft":1158,"via":"OpenAI","reqs":18},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-5.1-codex","name":"OpenAI: GPT-5.1-Codex","org":"OpenAI","slug":"gpt-5.1-codex","matchKey":"gpt51codex","blurb":"GPT-5.1-Codex is a specialized version of GPT-5.1 optimized for software engineering and coding workflows.","created":1763060298,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["text","image"],"out":["text"],"price":{"in":1.25,"out":10,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Azure","in":1.25,"out":10,"cacheRead":0.13,"uptime":100,"tps":19,"ttft":4866,"reqs":11}],"scores":{"arenaWebdev":{"v":1336.2804889632296,"ci":11.7,"rank":102,"votes":4861,"as":"gpt-5.1-codex","effort":""},"swebenchAgent":{"v":0.66,"as":"GPT 5.1 Codex","agent":"mini-SWE-agent","date":"2025-11-24","checked":false}},"perf":{"uptime":100,"cache":0.13,"n":1,"tps":19,"ttft":4866,"via":"Azure","reqs":11},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-5.2-chat","name":"OpenAI: GPT-5.2 Chat","org":"OpenAI","slug":"gpt-5.2-chat","matchKey":"gpt52chat","blurb":"GPT-5.2 Chat (AKA Instant) is the fast, lightweight member of the 5.2 family, optimized for low-latency chat while retaining strong general intelligence.","created":1765389783,"cutoff":null,"hf":null,"ctx":128000,"maxOut":32000,"in":["file","image","text"],"out":["text"],"price":{"in":1.75,"out":14,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":1.75,"out":14,"cacheRead":0.175}],"scores":{"arenaChat":{"v":1476.2585438549113,"ci":4.1,"rank":27,"votes":34176,"as":"gpt-5.2-chat-latest-20260210","effort":""},"arenaVision":{"v":1278.3379616392021,"ci":6.8,"rank":27,"votes":15432,"as":"gpt-5.2-chat-latest-20260210","effort":""}},"perf":{"cache":0.175,"n":1},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-5.2-codex","name":"OpenAI: GPT-5.2-Codex","org":"OpenAI","slug":"gpt-5.2-codex","matchKey":"gpt52codex","blurb":"GPT-5.2-Codex is an upgraded version of GPT-5.1-Codex optimized for software engineering and coding workflows.","created":1768409315,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["text","image"],"out":["text"],"price":{"in":1.75,"out":14,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"Azure","in":1.75,"out":14,"cacheRead":0.175,"uptime":100,"tps":48,"ttft":3559,"reqs":125}],"scores":{"arenaWebdev":{"v":1338.6977743667746,"ci":9,"rank":101,"votes":6402,"as":"gpt-5.2-codex","effort":""},"swebenchAgent":{"v":0.728,"as":"GPT 5.2 Codex","agent":"mini-SWE-agent","date":"2026-02-19","checked":false}},"perf":{"uptime":100,"cache":0.175,"n":1,"tps":48,"ttft":3559,"via":"Azure","reqs":125},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-image-1","name":"OpenAI: GPT Image 1","org":"OpenAI","slug":"gpt-image-1","matchKey":"gptimage1","blurb":"OpenAI's GPT Image 1 generates and edits images via the dedicated Images API.","created":1782264713,"cutoff":null,"hf":null,"ctx":400000,"maxOut":360000,"in":["text","image"],"out":["image"],"price":{"in":10,"out":10,"imageOut":0.00004,"webSearch":0.01},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":10,"out":10,"cacheRead":1.25,"uptime":99.7,"ttft":43957,"reqs":6}],"scores":{"arenaImageGen":{"v":1115.0750710002508,"ci":3.2,"rank":46,"votes":264897,"as":"gpt-image-1","effort":""},"arenaImageEdit":{"v":1138.887015649243,"ci":2.8,"rank":51,"votes":2853022,"as":"gpt-image-1","effort":""}},"perf":{"uptime":99.7,"cache":1.25,"n":1,"ttft":43957,"via":"OpenAI","reqs":6},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-image-1-mini","name":"OpenAI: GPT Image 1 Mini","org":"OpenAI","slug":"gpt-image-1-mini","matchKey":"gptimage1mini","blurb":"A cost-efficient variant of GPT Image 1 for high-quality image generation at reduced latency and cost via OpenAI's dedicated Images API.","created":1782264713,"cutoff":null,"hf":null,"ctx":400000,"maxOut":360000,"in":["text","image"],"out":["image"],"price":{"in":2.5,"out":2.5,"imageOut":0.000008,"webSearch":0.01},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":2.5,"out":2.5,"cacheRead":0.25,"uptime":100,"ttft":9560,"reqs":35}],"scores":{"arenaImageGen":{"v":1109.3699890439939,"ci":3.2,"rank":50,"votes":163843,"as":"gpt-image-1-mini","effort":""},"arenaImageEdit":{"v":1124.2728399511423,"ci":3,"rank":52,"votes":653148,"as":"gpt-image-1-mini","effort":""}},"perf":{"uptime":100,"cache":0.25,"n":1,"ttft":9560,"via":"OpenAI","reqs":35},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-image-2","name":"OpenAI: GPT Image 2","org":"OpenAI","slug":"gpt-image-2","matchKey":"gptimage2","blurb":"OpenAI's latest image generation model.","created":1782264714,"cutoff":null,"hf":null,"ctx":400000,"maxOut":360000,"in":["text","image"],"out":["image"],"price":{"in":8,"out":8,"imageOut":0.00003,"webSearch":0.01},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":8,"out":8,"cacheRead":2,"uptime":99.9,"ttft":31420,"reqs":844}],"scores":{"arenaImageGen":{"v":1381.1493148248587,"ci":4.4,"rank":3,"votes":78731,"as":"gpt-image-2 (medium)","effort":"medium"},"arenaImageEdit":{"v":1461.0473648794402,"ci":3.5,"rank":3,"votes":235928,"as":"gpt-image-2 (medium)","effort":"medium"}},"perf":{"uptime":99.9,"cache":2,"n":1,"ttft":31420,"via":"OpenAI","reqs":844},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-image-2.5-flare","name":"OpenAI: GPT Image 2.5 Flare","org":"OpenAI","slug":"gpt-image-2.5-flare","matchKey":"gptimage25flare","blurb":"GPT Image 2.5 Flare is an image generation and editing model from OpenAI, positioned as the speed-oriented tier of the GPT Image 2.5 series.","created":1788916350,"cutoff":null,"hf":null,"ctx":400000,"maxOut":360000,"in":["text","image"],"out":["image"],"price":{"in":8,"out":8,"imageOut":0.00003,"webSearch":0.01},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":8,"out":8,"cacheRead":2,"uptime":99.8,"ttft":12375,"reqs":723}],"scores":{"arenaImageGen":{"v":1398.9834153444895,"ci":12.7,"rank":2,"votes":2856,"as":"gpt-image-2.5-flare","effort":""},"arenaImageEdit":{"v":1490.632142005645,"ci":9.5,"rank":2,"votes":5676,"as":"gpt-image-2.5-flare","effort":""}},"perf":{"uptime":99.8,"cache":2,"n":1,"ttft":12375,"via":"OpenAI","reqs":723},"open":false,"rankable":true,"scored":2},{"id":"openai/gpt-image-2.5-sunburst","name":"OpenAI: GPT Image 2.5 Sunburst","org":"OpenAI","slug":"gpt-image-2.5-sunburst","matchKey":"gptimage25sunburst","blurb":"GPT Image 2.5 Sunburst is an image generation and editing model from OpenAI, positioned as the precision-oriented tier of the GPT Image 2.5 series.","created":1788916368,"cutoff":null,"hf":null,"ctx":400000,"maxOut":360000,"in":["text","image"],"out":["image"],"price":{"in":8,"out":8,"imageOut":0.00003,"webSearch":0.01},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":8,"out":8,"cacheRead":2,"uptime":97.1,"ttft":24454,"reqs":1485}],"scores":{"arenaImageGen":{"v":1420.7153252330818,"ci":13.1,"rank":1,"votes":3149,"as":"gpt-image-2.5-sunburst","effort":""},"arenaImageEdit":{"v":1520.443075793235,"ci":9.2,"rank":1,"votes":6704,"as":"gpt-image-2.5-sunburst","effort":""}},"perf":{"uptime":97.1,"cache":2,"n":1,"ttft":24454,"via":"OpenAI","reqs":1485},"open":false,"rankable":true,"scored":2},{"id":"rekaai/reka-flash-3","name":"Reka Flash 3","org":"rekaai","slug":"reka-flash-3","matchKey":"rekaflash3","blurb":"Reka Flash 3 is a general-purpose, instruction-tuned large language model with 21 billion parameters, developed by Reka.","created":1741812813,"cutoff":"2025-01-31","hf":"RekaAI/reka-flash-3","ctx":65536,"maxOut":58982,"in":["text"],"out":["text"],"price":{"in":0.1,"out":0.2},"caps":{"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Reka","quant":"fp8","in":0.1,"out":0.2,"uptime":100,"tps":91,"ttft":986,"reqs":19}],"scores":{"eqWriting":{"v":1225.5,"as":"RekaAI/reka-flash-3","version":"Creative Writing v3"},"eqLongform":{"v":41.2,"as":"RekaAI/reka-flash-3","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":1,"tps":91,"ttft":986,"via":"Reka","reqs":19},"open":true,"rankable":true,"scored":2},{"id":"x-ai/grok-imagine-image-2.0","name":"xAI: Grok Imagine Image 2.0","org":"xAI","slug":"grok-imagine-image-2.0","matchKey":"grokimagineimage20","blurb":"Grok Imagine Image 2.0 is an image generation and editing model from xAI.","created":1786486044,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000958083832335329,"imageIn":0.01},"caps":{"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"xAI","in":0,"out":0,"uptime":100,"ttft":15538,"reqs":269}],"scores":{"arenaImageGen":{"v":1315.065267643766,"ci":11.7,"rank":5,"votes":2681,"as":"grok-imagine-image-2.0 (low)","effort":"low"},"arenaImageEdit":{"v":1438.840391118557,"ci":8.5,"rank":4,"votes":5941,"as":"grok-imagine-image-2.0 (low)","effort":"low"}},"perf":{"uptime":100,"n":1,"ttft":15538,"via":"xAI","reqs":269},"open":false,"rankable":true,"scored":2},{"id":"deepseek/deepseek-v4.1-flash","name":"DeepSeek: DeepSeek V4.1 Flash","org":"DeepSeek","slug":"deepseek-v4.1-flash","matchKey":"deepseekv41flash","blurb":"DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture.","created":1789021285,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V4.1-Flash","ctx":1048576,"maxOut":384000,"in":["text","image"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"Morph","maxOut":943718,"in":0.12,"out":0.48,"cacheRead":0.0036,"uptime":97.7,"tps":6,"ttft":2501,"reqs":24168},{"name":"Relace","quant":"fp4","maxOut":943718,"in":0.13,"out":0.52,"cacheRead":0.01,"uptime":99.7,"noSchema":1,"tps":38,"ttft":1525,"reqs":72530},{"name":"DeepInfra","quant":"fp8","maxOut":131072,"in":0.14,"out":0.42,"cacheRead":0.0042,"uptime":99.7,"tps":42,"ttft":1693,"reqs":154915},{"name":"Sail Research","quant":"fp4","in":0.2,"out":0.6,"cacheRead":0.04,"uptime":99.9,"tps":91,"ttft":1051,"reqs":3625},{"name":"Wafer","maxOut":943718,"in":0.2,"out":0.6,"cacheRead":0.006,"uptime":99.8,"tps":55,"ttft":1412,"reqs":20073},{"name":"CoreWeave","quant":"fp8","maxOut":943718,"in":0.2,"out":0.65,"cacheRead":0.03,"uptime":99.6,"tps":208,"ttft":421,"reqs":8706},{"name":"StreamLake","quant":"fp8","ctx":1024000,"in":0.21969,"out":0.87876,"cacheRead":0.004394,"uptime":99.3,"noSchema":1,"tps":84,"ttft":1669,"reqs":11780},{"name":"Fireworks","maxOut":943718,"in":0.22,"out":0.66,"cacheRead":0.007,"uptime":99.6,"tps":128,"ttft":1120,"reqs":16575},{"name":"Phala","maxOut":393216,"in":0.276,"out":1.104,"cacheRead":0.00552,"uptime":99.2,"noSchema":1,"tps":60,"ttft":1225,"reqs":4021},{"name":"GMICloud","quant":"fp8","ctx":1048575,"maxOut":943717,"in":0.285,"out":1.14,"cacheRead":0.0057,"uptime":99.2,"noSchema":1,"tps":90,"ttft":1749,"reqs":11553},{"name":"Novita","quant":"fp8","maxOut":393216,"in":0.285,"out":1.14,"cacheRead":0.0057,"uptime":100,"noSchema":1,"tps":132,"ttft":1478,"reqs":63129},{"name":"AtlasCloud","quant":"fp8","maxOut":393216,"in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.3,"tps":99,"ttft":1042,"reqs":17552},{"name":"BaseTen","quant":"fp8","maxOut":32768,"in":0.3,"out":1.2,"cacheRead":0.03,"uptime":97.4,"noSchema":1},{"name":"Makora","maxOut":393216,"in":0.3,"out":1.2,"cacheRead":0.006,"uptime":98.3,"tps":56,"ttft":194,"reqs":26996},{"name":"DigitalOcean","maxOut":943718,"in":0.3,"out":1.2,"cacheRead":0.006,"uptime":98.3,"tps":29,"ttft":807,"reqs":3969},{"name":"Alibaba","ctx":1000000,"maxOut":393216,"in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.9,"noSchema":1,"tps":105,"ttft":854,"reqs":11735},{"name":"Together","maxOut":943718,"in":0.3,"out":1.2,"cacheRead":0.006,"uptime":99.9,"tps":153,"ttft":418,"reqs":143245},{"name":"SiliconFlow","quant":"fp8","maxOut":393216,"in":0.3,"out":1.2,"cacheRead":0.006,"uptime":99.5,"noSchema":1,"tps":87,"ttft":1468,"reqs":8409},{"name":"Modal","maxOut":943718,"in":0.3,"out":1.2,"cacheRead":0.03,"uptime":99.9,"tps":133,"ttft":661,"reqs":6452},{"name":"BaseTen","quant":"fp8","maxOut":32768,"in":0.3,"out":1.2,"cacheRead":0.007,"uptime":97.4,"noSchema":1,"tps":46,"ttft":2297,"reqs":1687,"region":"us"},{"name":"Parasail","quant":"fp8","maxOut":943718,"in":0.3,"out":1.2,"cacheRead":0.006,"uptime":91.1,"tps":70,"ttft":1694,"reqs":2544},{"name":"DeepSeek","in":0.3,"out":1.2,"cacheRead":0.006,"uptime":100,"noSchema":1,"tps":73,"ttft":876,"reqs":208343},{"name":"Venice","quant":"fp8","ctx":1000000,"maxOut":131072,"in":0.375,"out":1.5,"cacheRead":0.0075,"uptime":97.9,"tps":58,"ttft":10825,"reqs":6327}],"scores":{"arenaWebdev":{"v":1613.9516473561516,"ci":17.1,"rank":16,"votes":1361,"as":"deepseek-v4.1-flash-max","effort":"max"}},"perf":{"uptime":100,"cache":0.0036,"n":23,"tps":73,"ttft":876,"tpsBest":208,"via":"DeepSeek","reqs":208343,"regions":["us"]},"open":true,"rankable":true,"scored":1},{"id":"meta-llama/llama-3.3-70b-instruct","name":"Meta: Llama 3.3 70B Instruct","org":"Meta","slug":"llama-3.3-70b-instruct","matchKey":"llama3370binstruct","blurb":"The Meta Llama 3.3 multilingual large language model (LLM) is a pretrained and instruction tuned generative model in 70B (text in/text out).","created":1733506137,"cutoff":"2023-12-31","hf":"meta-llama/Llama-3.3-70B-Instruct","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.1,"out":0.32},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.1,"out":0.32,"uptime":98.8,"tps":12,"ttft":393,"reqs":27090},{"name":"Novita","quant":"bf16","ctx":12288,"maxOut":11059,"in":0.135,"out":0.4,"uptime":99.5,"noSchema":1,"tps":39,"ttft":1211,"reqs":5323},{"name":"AkashML","quant":"fp8","maxOut":128000,"in":0.2,"out":0.52,"cacheRead":0.1,"uptime":99.8,"tps":22,"ttft":1207,"reqs":10379},{"name":"Parasail","quant":"fp8","in":0.22,"out":0.5,"cacheRead":0.11,"uptime":99.4,"noTools":1,"tps":29,"ttft":668,"reqs":13950},{"name":"Cloudflare","quant":"fp8","ctx":24000,"maxOut":21600,"in":0.293,"out":2.253,"uptime":96.9,"noTools":1,"noSchema":1,"tps":36,"ttft":459,"reqs":1985},{"name":"SambaNova","maxOut":3072,"in":0.45,"out":0.9,"uptime":99,"noTools":1,"tps":80,"ttft":490,"reqs":1492},{"name":"Groq","maxOut":32768,"in":0.59,"out":0.79,"cacheRead":0.295,"uptime":99.7,"noSchema":1,"tps":211,"ttft":292,"reqs":7813},{"name":"CoreWeave","quant":"fp16","ctx":128000,"maxOut":115200,"in":0.71,"out":0.71,"cacheRead":0.71,"uptime":99.9,"tps":55,"ttft":707,"reqs":994},{"name":"Google","region":"us","ctx":128000,"maxOut":8192,"in":0.72,"out":0.72},{"name":"Google","ctx":128000,"maxOut":115200,"in":0.72,"out":0.72,"noTools":1,"noSchema":1,"tps":109,"ttft":2071,"reqs":1,"region":"us"},{"name":"Together","maxOut":2048,"in":1.04,"out":1.04,"uptime":93.4,"tps":30,"ttft":670,"reqs":335}],"scores":{"bfcl":{"v":0.319,"as":"Llama-3.3-70B-Instruct (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":99.9,"regions":["us"],"n":11,"tps":12,"ttft":393,"tpsBest":211,"via":"DeepInfra","reqs":27090},"open":true,"rankable":true,"scored":1},{"id":"google/gemini-2.5-pro-preview","name":"Google: Gemini 2.5 Pro Preview 06-05","org":"Google","slug":"gemini-2.5-pro-preview","matchKey":"gemini25propreview","blurb":"Gemini 2.5 Pro is Google’s state-of-the-art AI model designed for advanced reasoning, coding, mathematics, and scientific tasks.","created":1749137257,"cutoff":"2025-01-31","hf":null,"ctx":1048576,"maxOut":65536,"in":["file","image","text","audio"],"out":["text"],"price":{"in":1.25,"out":10,"imageIn":0.00000125,"audioIn":1.25,"webSearch":0.014},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Google","region":"global","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":95.9},{"name":"Google","in":2.25,"out":18,"cacheRead":0.225,"cacheWrite":0.675,"uptime":99.9},{"name":"Google AI Studio","in":0.625,"out":5,"cacheRead":0.0625,"cacheWrite":0.1875,"uptime":99.9},{"name":"Google","region":"eu","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":96.8},{"name":"Google AI Studio","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":98.3},{"name":"Google","region":"us","in":1.25,"out":10,"cacheRead":0.125,"cacheWrite":0.375,"uptime":83.7,"tps":90,"ttft":3301,"reqs":11},{"name":"Google AI Studio","in":2.25,"out":18,"cacheRead":0.225,"cacheWrite":0.675,"tps":83,"ttft":4079,"reqs":552}],"scores":{"gpqa":{"v":0.8484848484848485,"as":"Gemini 2.5 Pro Preview (Jun 2025)","effort":"","date":"2025-06-05","stderr":0.02554565042660364}},"perf":{"uptime":99.9,"cache":0.0625,"regions":["eu","global","us"],"n":7,"tps":82,"ttft":2444,"tpsBest":101,"via":"Google","reqs":12320},"open":false,"rankable":true,"scored":1},{"id":"qwen/qwen3.8-2.4t-a95b","name":"Qwen: Qwen3.8 2.4T A95B","org":"Qwen","slug":"qwen3.8-2.4t-a95b","matchKey":"qwen3824ta95b","blurb":"Qwen3.8 2.4T A95B is an open-weight sparse mixture-of-experts model from Qwen and the open-weight variant of [Qwen3.8 Max](/qwen/qwen3.8-max), with 95 billion active parameters out of 2.4 trillion total.","created":1786551702,"cutoff":null,"hf":"Qwen/Qwen3.8-2.4T-A95B","ctx":1048576,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":2,"out":6},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","medium","low"]},"moderated":false,"providers":[{"name":"Novita","ctx":1000000,"in":2,"out":6,"cacheRead":0.25,"uptime":99.9,"tps":36,"ttft":8170,"reqs":86},{"name":"Alibaba","ctx":1000000,"in":2,"out":6,"cacheRead":0.25,"cacheWrite":2.5,"uptime":99.9,"tps":33,"ttft":2725,"reqs":15},{"name":"SiliconFlow","quant":"fp8","in":2,"out":6,"cacheRead":0.25,"uptime":100,"tps":32,"ttft":1920,"reqs":694},{"name":"Venice","ctx":262144,"maxOut":65536,"in":2,"out":6,"cacheRead":0.25,"uptime":99.3,"noSchema":1,"tps":69,"ttft":724,"reqs":3},{"name":"Modal","ctx":1000000,"maxOut":262144,"in":2,"out":6,"cacheRead":0.25,"uptime":99.9,"tps":76,"ttft":828,"reqs":5},{"name":"DeepInfra","quant":"fp4","ctx":262144,"in":2,"out":6,"cacheRead":0.2,"uptime":100,"tps":28,"ttft":802,"reqs":4276},{"name":"Together","ctx":1010000,"maxOut":909000,"in":2,"out":6,"cacheRead":0.25,"uptime":100,"tps":124,"ttft":545,"reqs":7}],"scores":{"eqWriting":{"v":1840.9,"as":"Qwen/Qwen3.8-2.4T-A95B","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.25,"n":7,"tps":28,"ttft":802,"tpsBest":124,"via":"DeepInfra","reqs":4276},"open":true,"rankable":true,"scored":1},{"id":"google/gemini-2.5-flash-lite","name":"Google: Gemini 2.5 Flash Lite","org":"Google","slug":"gemini-2.5-flash-lite","matchKey":"gemini25flashlite","blurb":"Gemini 2.5 Flash-Lite is a lightweight reasoning model in the Gemini 2.5 family, optimized for ultra-low latency and cost efficiency.","created":1753200276,"cutoff":"2025-01-31","hf":null,"ctx":1048576,"maxOut":65535,"in":["text","image","file","audio","video"],"out":["text"],"price":{"in":0.1,"out":0.4,"imageIn":1e-7,"audioIn":0.3,"webSearch":0.014,"alt":{"batch":{"in":0.05,"out":0.2}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Google","region":"eu","in":0.1,"out":0.4,"cacheRead":0.01,"cacheWrite":0.083333,"uptime":99.8},{"name":"Google","in":0.1,"out":0.4,"cacheRead":0.01,"cacheWrite":0.083333,"uptime":94.5,"tps":15,"ttft":1042,"reqs":96974,"region":"eu"},{"name":"Google AI Studio","in":0.05,"out":0.2,"cacheRead":0.005,"cacheWrite":0.041667,"uptime":100},{"name":"Google AI Studio","in":0.1,"out":0.4,"cacheRead":0.01,"cacheWrite":0.083333,"uptime":99.7},{"name":"Google AI Studio","in":0.18,"out":0.72,"cacheRead":0.018,"cacheWrite":0.15,"uptime":100,"tps":254,"ttft":448,"reqs":37309}],"scores":{"bfcl":{"v":0.36869999999999997,"as":"Gemini-2.5-Flash-Lite (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":100,"cache":0.005,"regions":["eu"],"n":5,"tps":81,"ttft":441,"tpsBest":254,"via":"Google","reqs":476624},"open":false,"rankable":true,"scored":1},{"id":"deepseek/deepseek-v3.1-terminus","name":"DeepSeek: DeepSeek V3.1 Terminus","org":"DeepSeek","slug":"deepseek-v3.1-terminus","matchKey":"deepseekv31terminus","blurb":"DeepSeek-V3.1 Terminus is an update to [DeepSeek V3.1](/deepseek/deepseek-chat-v3.1) that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent ca","created":1758548275,"cutoff":"2025-03-31","hf":"deepseek-ai/DeepSeek-V3.1-Terminus","ctx":163840,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.27,"out":1},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"SiliconFlow","quant":"fp8","maxOut":147456,"in":0.27,"out":1,"uptime":97.5,"tps":21,"ttft":2079,"reqs":634},{"name":"Novita","quant":"fp8","ctx":131072,"in":0.27,"out":1,"cacheRead":0.135,"uptime":100,"tps":28,"ttft":2822,"reqs":548},{"name":"AtlasCloud","quant":"fp8","ctx":131072,"maxOut":65536,"in":0.3,"out":0.95,"cacheRead":0.13,"uptime":97.4,"tps":54,"ttft":1752,"reqs":186},{"name":"StreamLake","ctx":128000,"maxOut":32000,"in":0.3426,"out":1.0284,"uptime":96.6,"noTools":1,"tps":49,"ttft":1355,"reqs":266}],"scores":{"arenaChat":{"v":1417.4334590973172,"ci":10.1,"rank":122,"votes":3369,"as":"deepseek-v3.1-terminus-thinking","effort":"default"}},"perf":{"uptime":100,"n":4,"tps":21,"ttft":2079,"tpsBest":54,"via":"SiliconFlow","reqs":634},"open":true,"rankable":true,"scored":1},{"id":"meta/muse-glimmer-30b","name":"Meta: Muse Glimmer 30B","org":"Meta","slug":"muse-glimmer-30b","matchKey":"museglimmer30b","blurb":"Muse Glimmer 30B is a dense, open-weight multimodal model from Meta Superintelligence Labs, distilled from Muse Spark and optimized for autonomous agents on consumer hardware.","created":1786302394,"cutoff":null,"hf":"meta-models/Muse-Glimmer-30B","ctx":131072,"maxOut":16384,"in":["text","image"],"out":["text"],"price":{"in":0.3,"out":1.2,"alt":{"batch":{"in":0.175,"out":0.75}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"Phala","maxOut":117964,"in":0.3,"out":1.1,"cacheRead":0.04,"uptime":100,"tps":132,"ttft":551,"reqs":1395},{"name":"DeepInfra","quant":"bf16","in":0.3,"out":1.2,"cacheRead":0.04,"uptime":100,"tps":6,"ttft":1896,"reqs":770},{"name":"Fireworks","maxOut":117964,"in":0.35,"out":1.5,"cacheRead":0.04,"uptime":96.8,"tps":55,"ttft":843,"reqs":104},{"name":"Together","maxOut":117964,"in":0.35,"out":1.5,"cacheRead":0.04,"uptime":99.9,"tps":87,"ttft":285,"reqs":48}],"scores":{"eqWriting":{"v":1789.7,"as":"meta-models/Muse-Glimmer-30B","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.04,"n":4,"tps":132,"ttft":551,"via":"Phala","reqs":1395},"open":true,"rankable":true,"scored":1},{"id":"qwen/qwen3-coder-30b-a3b-instruct","name":"Qwen: Qwen3 Coder 30B A3B Instruct","org":"Qwen","slug":"qwen3-coder-30b-a3b-instruct","matchKey":"qwen3coder30ba3binstruct","blurb":"Qwen3-Coder-30B-A3B-Instruct is a 30.5B parameter Mixture-of-Experts (MoE) model with 128 experts (8 active per forward pass), designed for advanced code generation, repository-scale understanding, and agentic tool use.","created":1753972379,"cutoff":"2025-06-30","hf":"Qwen/Qwen3-Coder-30B-A3B-Instruct","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.07,"out":0.28},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Novita","quant":"fp8","ctx":160000,"maxOut":32768,"in":0.07,"out":0.27,"uptime":96.5,"noSchema":1,"tps":31,"ttft":1287,"reqs":683},{"name":"SiliconFlow","quant":"fp8","in":0.07,"out":0.28,"uptime":97.8,"tps":26,"ttft":1390,"reqs":434},{"name":"Amazon Bedrock","in":0.15,"out":0.6,"noSchema":1},{"name":"Alibaba","maxOut":65536,"in":0.2925,"out":1.4625,"uptime":100,"noSchema":1,"tps":36,"ttft":1010,"reqs":16}],"scores":{"swebenchAgent":{"v":0.604,"as":"Qwen3-Coder-30B-A3B-Instruct","agent":"EntroPO + R2E","date":"2025-09-01","checked":true}},"perf":{"uptime":100,"n":4,"tps":31,"ttft":1287,"tpsBest":36,"via":"Novita","reqs":683},"open":true,"rankable":true,"scored":1},{"id":"x-ai/grok-4.7","name":"SpaceXAI: Grok 4.7","org":"SpaceXAI","slug":"grok-4.7","matchKey":"grok47","blurb":"Grok 4.7 is SpaceXAI's flagship model for coding, agentic tasks, and knowledge work, succeeding Grok 4.6.","created":1790007541,"cutoff":null,"hf":null,"ctx":500000,"maxOut":450000,"in":["text","image","file"],"out":["text"],"price":{"in":1.6,"out":4.8,"webSearch":0.005},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"xAI","in":1.6,"out":4.8,"cacheRead":0.4,"uptime":99.5},{"name":"xAI","in":1.6,"out":4.8,"cacheRead":0.4,"uptime":99.1},{"name":"xAI","in":3.2,"out":9.6,"cacheRead":0.8,"uptime":99.7},{"name":"xAI","in":3.2,"out":9.6,"cacheRead":0.8,"uptime":99.5,"tps":62,"ttft":1437,"reqs":8386}],"scores":{"terminal":{"v":0.37579999999999997,"as":"Grok 4.7","agent":"Grok Build","effort":"xhigh","ci":0.0354,"trials":330,"cost":3683.287146,"secs":5700.9,"date":"2026-09-21"}},"perf":{"uptime":99.7,"cache":0.4,"n":4,"tps":62,"ttft":1437,"via":"xAI","reqs":8386},"open":false,"rankable":true,"scored":1},{"id":"mistralai/ministral-14b-2512","name":"Mistral: Ministral 3 14B 2512","org":"Mistral","slug":"ministral-14b-2512","matchKey":"ministral14b2512","blurb":"The largest model in the Ministral 3 family, Ministral 3 14B offers frontier capabilities and performance comparable to its larger Mistral Small 3.2 24B counterpart.","created":1764681735,"cutoff":null,"hf":"mistralai/Ministral-3-14B-Instruct-2512","ctx":262144,"maxOut":209715,"in":["text","image"],"out":["text"],"price":{"in":0.2,"out":0.2},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.2,"out":0.2,"cacheRead":0.02,"uptime":99},{"name":"Mistral","in":0.2,"out":0.2,"cacheRead":0.02,"uptime":99.4},{"name":"Mistral","region":"eu","in":0.22,"out":0.22,"cacheRead":0.022,"uptime":100,"tps":52,"ttft":318,"reqs":1053}],"scores":{"eqLongform":{"v":31.9,"as":"mistralai/Ministral-3-14B-Instruct-2512","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.02,"regions":["eu"],"n":3,"tps":52,"ttft":318,"via":"Mistral","reqs":1053},"open":true,"rankable":true,"scored":1},{"id":"mistralai/mistral-medium-3","name":"Mistral: Mistral Medium 3","org":"Mistral","slug":"mistral-medium-3","matchKey":"mistralmedium3","blurb":"Mistral Medium 3 is a high-performance enterprise-grade language model designed to deliver frontier-level capabilities at significantly reduced operational cost.","created":1746627341,"cutoff":"2025-03-31","hf":null,"ctx":131072,"maxOut":104857,"in":["text","image","file"],"out":["text"],"price":{"in":0.4,"out":2},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.4,"out":2,"cacheRead":0.04,"uptime":99.8},{"name":"Mistral","in":0.4,"out":2,"cacheRead":0.04,"uptime":99.8},{"name":"Mistral","region":"eu","in":0.44,"out":2.2,"cacheRead":0.044,"tps":110,"ttft":434,"reqs":55}],"scores":{"bfcl":{"v":0.37560000000000004,"as":"Mistral-Medium-2505 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":99.8,"cache":0.04,"regions":["eu"],"n":3,"tps":110,"ttft":434,"via":"Mistral","reqs":55},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-4o-mini","name":"OpenAI: GPT-4o-mini","org":"OpenAI","slug":"gpt-4o-mini","matchKey":"gpt4omini","blurb":"GPT-4o mini is OpenAI's newest model after [GPT-4 Omni](/models/openai/gpt-4o), supporting both text and image inputs with text outputs.","created":1721260800,"cutoff":"2023-10-31","hf":null,"ctx":128000,"maxOut":16384,"in":["text","image","file"],"out":["text"],"price":{"in":0.15,"out":0.6,"alt":{"batch":{"in":0.075,"out":0.3}}},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","in":0.15,"out":0.6,"cacheRead":0.075,"uptime":99.8,"noTools":1},{"name":"OpenAI","in":0.15,"out":0.6,"cacheRead":0.075,"uptime":99.9,"tps":52,"ttft":543,"reqs":114979},{"name":"Azure","region":"eu","in":0.165,"out":0.66,"cacheRead":0.0825,"uptime":100,"noTools":1,"tps":42,"ttft":809,"reqs":3}],"scores":{"eqWriting":{"v":871,"as":"gpt-4o-mini","version":"Creative Writing v3"}},"perf":{"uptime":100,"cache":0.075,"regions":["eu"],"n":3,"tps":52,"ttft":543,"via":"OpenAI","reqs":114979},"open":false,"rankable":true,"scored":1},{"id":"qwen/qwen3-235b-a22b-thinking-2507","name":"Qwen: Qwen3 235B A22B Thinking 2507","org":"Qwen","slug":"qwen3-235b-a22b-thinking-2507","matchKey":"qwen3235ba22bthinking2507","blurb":"Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language model optimized for complex reasoning tasks.","created":1753449557,"cutoff":"2025-06-30","hf":"Qwen/Qwen3-235B-A22B-Thinking-2507","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.23,"out":2.3},"caps":{"tools":true,"json":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.23,"out":2.3,"uptime":100,"tps":79,"ttft":325,"reqs":36},{"name":"Novita","quant":"fp8","maxOut":32768,"in":0.3,"out":3,"uptime":98.6,"tps":11,"ttft":1166,"reqs":15},{"name":"Venice","quant":"fp8","ctx":128000,"maxOut":16384,"in":0.45,"out":3.5,"uptime":91.3}],"scores":{"arenaChat":{"v":1400.0554269590039,"ci":6.6,"rank":149,"votes":8796,"as":"qwen3-235b-a22b-thinking-2507","effort":"default"}},"perf":{"uptime":100,"n":3,"tps":79,"ttft":325,"via":"Alibaba","reqs":36},"open":true,"rankable":true,"scored":1},{"id":"z-ai/glm-4.5-air","name":"Z.ai: GLM 4.5 Air","org":"Z.ai","slug":"glm-4.5-air","matchKey":"glm45air","blurb":"GLM-4.5-Air is the lightweight variant of our latest flagship model family, also purpose-built for agent-centric applications.","created":1753471258,"cutoff":"2024-12-31","hf":"zai-org/GLM-4.5-Air","ctx":131072,"maxOut":98304,"in":["text"],"out":["text"],"price":{"in":0.13,"out":0.85},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"bf16","in":0.13,"out":0.85,"cacheRead":0.025,"uptime":99.8,"tps":51,"ttft":1172,"reqs":2148},{"name":"SiliconFlow","quant":"fp8","maxOut":117964,"in":0.14,"out":0.86,"uptime":100,"tps":29,"ttft":1371,"reqs":115},{"name":"Z.AI","quant":"fp8","in":0.2,"out":1.1,"cacheRead":0.03,"uptime":98.8,"tps":45,"ttft":1569,"reqs":418}],"scores":{"arenaChat":{"v":1373.2072994769237,"ci":4.3,"rank":182,"votes":30367,"as":"glm-4.5-air","effort":""}},"perf":{"uptime":100,"cache":0.025,"n":3,"tps":51,"ttft":1172,"via":"Novita","reqs":2148},"open":true,"rankable":true,"scored":1},{"id":"amazon/nova-micro-v1","name":"Amazon: Nova Micro 1.0","org":"Amazon","slug":"nova-micro-v1","matchKey":"novamicrov1","blurb":"Amazon Nova Micro 1.0 is a text-only model that delivers the lowest latency responses in the Amazon Nova family of models at a very low cost.","created":1733437237,"cutoff":"2024-10-31","hf":null,"ctx":128000,"maxOut":5120,"in":["text"],"out":["text"],"price":{"in":0.035,"out":0.14},"caps":{"tools":true},"reasoning":null,"moderated":true,"providers":[{"name":"Amazon Bedrock","region":"eu","in":0.035,"out":0.14,"uptime":74.9},{"name":"Amazon Bedrock","in":0.035,"out":0.14,"uptime":98.3,"tps":52,"ttft":389,"reqs":5807,"region":"eu"}],"scores":{"bfcl":{"v":0.2229,"as":"Amazon-Nova-Micro-v1:0 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":98.3,"regions":["eu"],"n":2,"tps":67,"ttft":362,"via":"Amazon Bedrock","reqs":6738},"open":false,"rankable":true,"scored":1},{"id":"amazon/nova-pro-v1","name":"Amazon: Nova Pro 1.0","org":"Amazon","slug":"nova-pro-v1","matchKey":"novaprov1","blurb":"Amazon Nova Pro 1.0 is a capable multimodal model from Amazon focused on providing a combination of accuracy, speed, and cost for a wide range of tasks.","created":1733436303,"cutoff":"2024-10-31","hf":null,"ctx":300000,"maxOut":5120,"in":["text","image"],"out":["text"],"price":{"in":0.8,"out":3.2},"caps":{"tools":true},"reasoning":null,"moderated":true,"providers":[{"name":"Amazon Bedrock","region":"eu","in":0.8,"out":3.2,"uptime":100},{"name":"Amazon Bedrock","in":0.8,"out":3.2,"uptime":100,"tps":159,"ttft":362,"reqs":35,"region":"eu"}],"scores":{"bfcl":{"v":0.24969999999999998,"as":"Amazon-Nova-Pro-v1:0 (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":100,"regions":["eu"],"n":2,"tps":173,"ttft":548,"via":"Amazon Bedrock","reqs":40},"open":false,"rankable":true,"scored":1},{"id":"meta-llama/llama-3.1-70b-instruct","name":"Meta: Llama 3.1 70B Instruct","org":"Meta","slug":"llama-3.1-70b-instruct","matchKey":"llama3170binstruct","blurb":"Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors.","created":1721692800,"cutoff":"2023-12-31","hf":"meta-llama/Meta-Llama-3.1-70B-Instruct","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.4,"out":0.4},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.4,"out":0.4,"uptime":96.2,"tps":35,"ttft":185,"reqs":1371},{"name":"Amazon Bedrock","maxOut":8192,"in":0.72,"out":0.72,"uptime":100,"noTools":1,"noSchema":1,"tps":48,"ttft":448,"reqs":4287}],"scores":{"eqWriting":{"v":781.9,"as":"meta-llama/llama-3.1-70b-instruct","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":2,"tps":48,"ttft":448,"via":"Amazon Bedrock","reqs":4287},"open":true,"rankable":true,"scored":1},{"id":"minimax/minimax-m1","name":"MiniMax: MiniMax M1","org":"MiniMax","slug":"minimax-m1","matchKey":"minimaxm1","blurb":"MiniMax-M1 is a large-scale, open-weight reasoning model designed for extended context and high-efficiency inference.","created":1750200414,"cutoff":"2024-06-30","hf":null,"ctx":1000000,"maxOut":40000,"in":["text"],"out":["text"],"price":{"in":0.4,"out":2.2},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Minimax","in":0.4,"out":2.2,"uptime":100,"noTools":1,"tps":24,"ttft":779,"reqs":8},{"name":"Novita","quant":"bf16","in":0.55,"out":2.2,"uptime":100,"tps":16,"ttft":966,"reqs":3}],"scores":{"arenaChat":{"v":1363.9152089057316,"ci":4.3,"rank":189,"votes":34030,"as":"minimax-m1","effort":""}},"perf":{"uptime":100,"n":2,"tps":24,"ttft":779,"via":"Minimax","reqs":8},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-4o-2024-05-13","name":"OpenAI: GPT-4o (2024-05-13)","org":"OpenAI","slug":"gpt-4o-2024-05-13","matchKey":"gpt4o20240513","blurb":"GPT-4o (\"o\" for \"omni\") is OpenAI's latest AI model, supporting both text and image inputs with text outputs.","created":1715558400,"cutoff":"2023-10-31","hf":null,"ctx":128000,"maxOut":4096,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":15},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","in":5,"out":15,"uptime":96.1,"tps":36,"ttft":1898,"reqs":16},{"name":"OpenAI","in":5,"out":15,"uptime":100,"tps":83,"ttft":1100,"reqs":355}],"scores":{"arenaVision":{"v":1162.1531539529801,"ci":8.7,"rank":95,"votes":23273,"as":"gpt-4o-2024-05-13","effort":""}},"perf":{"uptime":100,"n":2,"tps":83,"ttft":1100,"via":"OpenAI","reqs":355},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-4o-2024-08-06","name":"OpenAI: GPT-4o (2024-08-06)","org":"OpenAI","slug":"gpt-4o-2024-08-06","matchKey":"gpt4o20240806","blurb":"The 2024-08-06 version of GPT-4o offers improved performance in structured outputs, with the ability to supply a JSON schema in the respone_format.","created":1722902400,"cutoff":"2023-10-31","hf":null,"ctx":128000,"maxOut":16384,"in":["text","image","file"],"out":["text"],"price":{"in":2.5,"out":10},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Azure","in":2.5,"out":10,"cacheRead":1.25,"uptime":99.9,"tps":63,"ttft":986,"reqs":224},{"name":"OpenAI","in":2.5,"out":10,"cacheRead":1.25,"uptime":99.9,"tps":27,"ttft":646,"reqs":733}],"scores":{"arenaVision":{"v":1118.9355912916667,"ci":12.3,"rank":118,"votes":3376,"as":"gpt-4o-2024-08-06","effort":""}},"perf":{"uptime":99.9,"cache":1.25,"n":2,"tps":27,"ttft":646,"tpsBest":63,"via":"OpenAI","reqs":733},"open":false,"rankable":true,"scored":1},{"id":"qwen/qwen3-30b-a3b","name":"Qwen: Qwen3 30B A3B","org":"Qwen","slug":"qwen3-30b-a3b","matchKey":"qwen330ba3b","blurb":"Qwen3, the latest generation in the Qwen large language model series, features both dense and mixture-of-experts (MoE) architectures to excel in reasoning, multilingual support, and advanced agent tasks.","created":1745878604,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-30B-A3B","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.12,"out":0.5},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","ctx":40960,"in":0.12,"out":0.5,"uptime":99.9,"tps":60,"ttft":900,"reqs":502},{"name":"Alibaba","maxOut":8192,"in":0.13,"out":0.52,"uptime":100,"noSchema":1,"tps":89,"ttft":404,"reqs":435}],"scores":{"eqLongform":{"v":35.6,"as":"qwen/qwen3-30b-a3b","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"n":2,"tps":60,"ttft":900,"tpsBest":89,"via":"DeepInfra","reqs":502},"open":true,"rankable":true,"scored":1},{"id":"qwen/qwen3-next-80b-a3b-thinking","name":"Qwen: Qwen3 Next 80B A3B Thinking","org":"Qwen","slug":"qwen3-next-80b-a3b-thinking","matchKey":"qwen3next80ba3bthinking","blurb":"Qwen3-Next-80B-A3B-Thinking is a reasoning-first chat model in the Qwen3-Next line that outputs structured “thinking” traces by default.","created":1757612284,"cutoff":"2025-09-30","hf":"Qwen/Qwen3-Next-80B-A3B-Thinking","ctx":262144,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.15,"out":1.2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Google","region":"global","maxOut":235929,"in":0.15,"out":1.2,"uptime":99.9,"tps":28,"ttft":442,"reqs":13},{"name":"Alibaba","ctx":131072,"in":0.15,"out":1.2,"uptime":100,"noSchema":1,"tps":197,"ttft":523,"reqs":20}],"scores":{"arenaChat":{"v":1368.9618369723216,"ci":5.9,"rank":184,"votes":13380,"as":"qwen3-next-80b-a3b-thinking","effort":"default"}},"perf":{"uptime":100,"regions":["global"],"n":2,"tps":197,"ttft":523,"via":"Alibaba","reqs":20},"open":true,"rankable":true,"scored":1},{"id":"inception/mercury-2","name":"Inception: Mercury 2","org":"Inception","slug":"mercury-2","matchKey":"mercury2","blurb":"Mercury 2 is an extremely fast reasoning LLM, and the first reasoning diffusion LLM (dLLM).","created":1772636275,"cutoff":null,"hf":null,"ctx":128000,"maxOut":50000,"in":["text"],"out":["text"],"price":{"in":0.25,"out":0.75},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","none"]},"moderated":false,"providers":[{"name":"Inception","in":0.25,"out":0.75,"cacheRead":0.025,"uptime":100,"tps":180,"ttft":468,"reqs":2967}],"scores":{"arenaWebdev":{"v":1166.5930317649709,"ci":25.2,"rank":125,"votes":904,"as":"mercury-2","effort":""}},"perf":{"uptime":100,"cache":0.025,"n":1,"tps":180,"ttft":468,"via":"Inception","reqs":2967},"open":false,"rankable":true,"scored":1},{"id":"krea/krea-2-large","name":"Krea: Krea 2 Large","org":"Krea","slug":"krea-2-large","matchKey":"krea2large","blurb":"Krea 2 Large is Krea's high-capability image generation model, more than twice the size of Krea 2 Medium.","created":1784574931,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000143712574850299},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Krea","in":0,"out":0,"uptime":100,"ttft":37033,"reqs":19}],"scores":{"arenaImageGen":{"v":1108.262138702522,"ci":4.5,"rank":51,"votes":39614,"as":"krea-2-large","effort":""}},"perf":{"uptime":100,"n":1,"ttft":37033,"via":"Krea","reqs":19},"open":false,"rankable":true,"scored":1},{"id":"krea/krea-2-medium","name":"Krea: Krea 2 Medium","org":"Krea","slug":"krea-2-medium","matchKey":"krea2medium","blurb":"Krea 2 Medium is Krea's balanced, cost-efficient image generation model and a practical starting point for a broad range of use cases.","created":1784574928,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000718562874251497},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Krea","in":0,"out":0,"uptime":100,"ttft":9512,"reqs":1}],"scores":{"arenaImageGen":{"v":1122.5803963744195,"ci":4.5,"rank":42,"votes":40205,"as":"krea-2-medium","effort":"medium"}},"perf":{"uptime":100,"n":1,"ttft":9512,"via":"Krea","reqs":1},"open":false,"rankable":true,"scored":1},{"id":"microsoft/phi-4","name":"Microsoft: Phi 4","org":"Microsoft","slug":"phi-4","matchKey":"phi4","blurb":"[Microsoft Research](/microsoft) Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed.","created":1736489872,"cutoff":"2024-06-30","hf":"microsoft/phi-4","ctx":16384,"maxOut":14745,"in":["text"],"out":["text"],"price":{"in":0.07,"out":0.14},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":0.07,"out":0.14,"uptime":100,"tps":29,"ttft":328,"reqs":500}],"scores":{"bfcl":{"v":0.2879,"as":"Phi-4 (Prompt)","effort":"","version":"V4","agent":"BFCL V4 · prompt-based tool calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":100,"n":1,"tps":29,"ttft":328,"via":"DeepInfra","reqs":500},"open":true,"rankable":true,"scored":1},{"id":"minimax/minimax-m2-her","name":"MiniMax: MiniMax M2-her","org":"MiniMax","slug":"minimax-m2-her","matchKey":"minimaxm2her","blurb":"MiniMax M2-her is a dialogue-first large language model built for immersive roleplay, character-driven chat, and expressive multi-turn conversations.","created":1769177239,"cutoff":null,"hf":null,"ctx":65536,"maxOut":2048,"in":["text"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Minimax","in":0.3,"out":1.2,"cacheRead":0.03,"uptime":100,"tps":26,"ttft":693,"reqs":388}],"scores":{"eqLongform":{"v":34.8,"as":"minimax-m2-her","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.03,"n":1,"tps":26,"ttft":693,"via":"Minimax","reqs":388},"open":false,"rankable":true,"scored":1},{"id":"mistralai/mistral-small-24b-instruct-2501","name":"Mistral: Mistral Small 3","org":"Mistral","slug":"mistral-small-24b-instruct-2501","matchKey":"mistralsmall24binstruct2501","blurb":"Mistral Small 3 is a 24B-parameter language model optimized for low-latency performance across common AI tasks.","created":1738255409,"cutoff":"2023-10-31","hf":"mistralai/Mistral-Small-24B-Instruct-2501","ctx":32768,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.05,"out":0.08},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.05,"out":0.08,"uptime":100,"tps":42,"ttft":325,"reqs":9899}],"scores":{"eqWriting":{"v":704.3,"as":"mistralai/Mistral-Small-24B-Instruct-2501","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":1,"tps":42,"ttft":325,"via":"DeepInfra","reqs":9899},"open":true,"rankable":true,"scored":1},{"id":"openai/gpt-3.5-turbo-0613","name":"OpenAI: GPT-3.5 Turbo (older v0613)","org":"OpenAI","slug":"gpt-3.5-turbo-0613","matchKey":"gpt35turbo0613","blurb":"GPT-3.5 Turbo is OpenAI's fastest model.","created":1706140800,"cutoff":"2021-09-30","hf":null,"ctx":4095,"maxOut":3685,"in":["text"],"out":["text"],"price":{"in":1,"out":2},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":1,"out":2,"uptime":100,"tps":8,"ttft":876,"reqs":8}],"scores":{"eqWriting":{"v":449.9,"as":"openai/gpt-3.5-turbo-0613","version":"Creative Writing v3"}},"perf":{"uptime":100,"n":1,"tps":8,"ttft":876,"via":"Azure","reqs":8},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-4o-mini-2024-07-18","name":"OpenAI: GPT-4o-mini (2024-07-18)","org":"OpenAI","slug":"gpt-4o-mini-2024-07-18","matchKey":"gpt4omini20240718","blurb":"GPT-4o mini is OpenAI's newest model after [GPT-4 Omni](/models/openai/gpt-4o), supporting both text and image inputs with text outputs.","created":1721260800,"cutoff":"2023-10-31","hf":null,"ctx":128000,"maxOut":16384,"in":["text","image","file"],"out":["text"],"price":{"in":0.15,"out":0.6},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":0.15,"out":0.6,"cacheRead":0.075,"uptime":100,"tps":18,"ttft":540,"reqs":9364}],"scores":{"arenaVision":{"v":1097.6008127263178,"ci":8.2,"rank":122,"votes":17341,"as":"gpt-4o-mini-2024-07-18","effort":""}},"perf":{"uptime":100,"cache":0.075,"n":1,"tps":18,"ttft":540,"via":"OpenAI","reqs":9364},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-5-pro","name":"OpenAI: GPT-5 Pro","org":"OpenAI","slug":"gpt-5-pro","matchKey":"gpt5pro","blurb":"GPT-5 Pro is OpenAI’s most advanced model, offering major improvements in reasoning, code quality, and user experience.","created":1759776663,"cutoff":"2024-09-30","hf":null,"ctx":400000,"maxOut":128000,"in":["image","text","file"],"out":["text"],"price":{"in":15,"out":120,"webSearch":0.01,"alt":{"batch":{"in":7.5,"out":60}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high"]},"moderated":true,"providers":[{"name":"OpenAI","in":15,"out":120,"uptime":100,"tps":2,"ttft":14963,"reqs":4}],"scores":{"frontierMath":{"v":0.5578947368421052,"as":"GPT-5 Pro","effort":"","date":"2026-06-12","stderr":0.029469977528223976}},"perf":{"uptime":100,"n":1,"tps":2,"ttft":14963,"via":"OpenAI","reqs":4},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-5.1-codex-mini","name":"OpenAI: GPT-5.1-Codex-Mini","org":"OpenAI","slug":"gpt-5.1-codex-mini","matchKey":"gpt51codexmini","blurb":"GPT-5.1-Codex-Mini is a smaller and faster version of GPT-5.1-Codex","created":1763057820,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["image","text"],"out":["text"],"price":{"in":0.25,"out":2,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Azure","in":0.25,"out":2,"cacheRead":0.03,"uptime":100,"tps":76,"ttft":4465,"reqs":70}],"scores":{"arenaWebdev":{"v":1243.9286259282542,"ci":22.3,"rank":116,"votes":1133,"as":"gpt-5.1-codex-mini","effort":""}},"perf":{"uptime":100,"cache":0.03,"n":1,"tps":76,"ttft":4465,"via":"Azure","reqs":70},"open":false,"rankable":true,"scored":1},{"id":"openai/gpt-5.2-pro","name":"OpenAI: GPT-5.2 Pro","org":"OpenAI","slug":"gpt-5.2-pro","matchKey":"gpt52pro","blurb":"GPT-5.2 Pro is OpenAI’s most advanced model, offering major improvements in agentic coding and long context performance over GPT-5 Pro.","created":1765389780,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["image","text","file"],"out":["text"],"price":{"in":21,"out":168,"webSearch":0.01,"alt":{"batch":{"in":10.5,"out":84}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium"]},"moderated":true,"providers":[{"name":"OpenAI","in":21,"out":168,"uptime":100,"tps":2,"ttft":4335,"reqs":24}],"scores":{"frontierMath":{"v":0.74,"as":"GPT-5.2 Pro","effort":"","date":"2026-06-13","stderr":0.026349537704451795}},"perf":{"uptime":100,"n":1,"tps":2,"ttft":4335,"via":"OpenAI","reqs":24},"open":false,"rankable":true,"scored":1},{"id":"openai/o3-mini-high","name":"OpenAI: o3 Mini High","org":"OpenAI","slug":"o3-mini-high","matchKey":"o3minihigh","blurb":"OpenAI o3-mini-high is the same model as [o3-mini](/openai/o3-mini) with reasoning_effort set to high.","created":1739372611,"cutoff":"2023-10-31","hf":null,"ctx":200000,"maxOut":100000,"in":["text","file"],"out":["text"],"price":{"in":1.1,"out":4.4,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high"]},"moderated":true,"providers":[{"name":"OpenAI","in":1.1,"out":4.4,"cacheRead":0.55,"uptime":100,"tps":142,"ttft":15007,"reqs":2}],"scores":{"arenaChat":{"v":1363.4599106805713,"ci":5.3,"rank":191,"votes":18589,"as":"o3-mini-high","effort":"high"}},"perf":{"uptime":100,"cache":0.55,"n":1,"tps":142,"ttft":15007,"via":"OpenAI","reqs":2},"open":false,"rankable":true,"scored":1},{"id":"openai/sora-2-pro","name":"OpenAI: Sora 2 Pro","org":"OpenAI","slug":"sora-2-pro","matchKey":"sora2pro","blurb":"OpenAI's flagship video generation model, delivering production-quality video with physics-accurate motion, synchronized audio, and world-state persistence across shots.","created":1774277521,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"logprobs":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":0,"out":0}],"scores":{"arenaVideoGen":{"v":1366.5517880630687,"ci":6.9,"rank":11,"votes":50415,"as":"sora-2-pro","effort":""}},"perf":{"n":1},"open":false,"rankable":true,"scored":1},{"id":"perplexity/sonar-pro","name":"Perplexity: Sonar Pro","org":"Perplexity","slug":"sonar-pro","matchKey":"sonarpro","blurb":"Note: Sonar Pro pricing includes Perplexity search pricing.","created":1741312423,"cutoff":null,"hf":null,"ctx":200000,"maxOut":8000,"in":["text","image"],"out":["text"],"price":{"in":3,"out":15,"webSearch":0.005},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Perplexity","in":3,"out":15,"uptime":100,"tps":72,"ttft":2136,"reqs":1161}],"scores":{"arenaSearch":{"v":1130.2067048282445,"ci":5.7,"rank":31,"votes":28519,"as":"ppl-sonar-pro-high","effort":"high"}},"perf":{"uptime":100,"n":1,"tps":72,"ttft":2136,"via":"Perplexity","reqs":1161},"open":false,"rankable":true,"scored":1},{"id":"perplexity/sonar-reasoning-pro","name":"Perplexity: Sonar Reasoning Pro","org":"Perplexity","slug":"sonar-reasoning-pro","matchKey":"sonarreasoningpro","blurb":"Note: Sonar Pro pricing includes Perplexity search pricing.","created":1741313308,"cutoff":null,"hf":null,"ctx":128000,"maxOut":115200,"in":["text","image"],"out":["text"],"price":{"in":2,"out":8,"webSearch":0.005},"caps":{"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Perplexity","in":2,"out":8,"uptime":100,"tps":25,"ttft":40331,"reqs":62}],"scores":{"arenaSearch":{"v":1138.6014177151164,"ci":5.7,"rank":29,"votes":29055,"as":"ppl-sonar-reasoning-pro-high","effort":"high"}},"perf":{"uptime":100,"n":1,"tps":25,"ttft":40331,"via":"Perplexity","reqs":62},"open":false,"rankable":true,"scored":1},{"id":"qwen/qwen-2.5-coder-32b-instruct","name":"Qwen2.5 Coder 32B Instruct","org":"qwen","slug":"qwen-2.5-coder-32b-instruct","matchKey":"qwen25coder32binstruct","blurb":"Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen).","created":1731368400,"cutoff":"2024-06-30","hf":"Qwen/Qwen2.5-Coder-32B-Instruct","ctx":32768,"maxOut":29491,"in":["text"],"out":["text"],"price":{"in":0.66,"out":1},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Cloudflare","in":0.66,"out":1,"uptime":100,"tps":27,"ttft":2103,"reqs":8264}],"scores":{"swebenchAgent":{"v":0.38,"as":"Qwen2.5 Coder 32B Instruct","agent":"Skywork-SWE-32B","date":"2025-06-16","checked":true}},"perf":{"uptime":100,"n":1,"tps":27,"ttft":2103,"via":"Cloudflare","reqs":8264},"open":true,"rankable":true,"scored":1},{"id":"qwen/qwen2.5-vl-72b-instruct","name":"Qwen: Qwen2.5 VL 72B Instruct","org":"Qwen","slug":"qwen2.5-vl-72b-instruct","matchKey":"qwen25vl72binstruct","blurb":"Qwen2.5-VL is proficient in recognizing common objects such as flowers, birds, fish, and insects.","created":1738410311,"cutoff":"2024-06-30","hf":"Qwen/Qwen2.5-VL-72B-Instruct","ctx":128000,"maxOut":115200,"in":["text","image"],"out":["text"],"price":{"in":0.8,"out":1},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"fp8","in":0.8,"out":1,"cacheRead":0.4,"uptime":100,"tps":29,"ttft":905,"reqs":1702}],"scores":{"arenaVision":{"v":1121.486301474211,"ci":10.8,"rank":116,"votes":3768,"as":"qwen2.5-vl-72b-instruct","effort":""}},"perf":{"uptime":100,"cache":0.4,"n":1,"tps":29,"ttft":905,"via":"Parasail","reqs":1702},"open":true,"rankable":true,"scored":1},{"id":"qwen/qwen3-235b-a22b","name":"Qwen: Qwen3 235B A22B","org":"Qwen","slug":"qwen3-235b-a22b","matchKey":"qwen3235ba22b","blurb":"Qwen3-235B-A22B is a 235B parameter mixture-of-experts (MoE) model developed by Qwen, activating 22B parameters per forward pass.","created":1745875757,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-235B-A22B","ctx":131072,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":0.455,"out":1.82},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.455,"out":1.82,"uptime":100,"tps":39,"ttft":695,"reqs":451}],"scores":{"arenaChat":{"v":1423.0293682990673,"ci":2.6,"rank":114,"votes":95670,"as":"qwen3-235b-a22b-instruct-2507","effort":""}},"perf":{"uptime":100,"n":1,"tps":39,"ttft":695,"via":"Alibaba","reqs":451},"open":true,"rankable":true,"scored":1},{"id":"qwen/qwen3-8b","name":"Qwen: Qwen3 8B","org":"Qwen","slug":"qwen3-8b","matchKey":"qwen38b","blurb":"Qwen3-8B is a dense 8.2B parameter causal language model from the Qwen3 series, designed for both reasoning-heavy tasks and efficient dialogue.","created":1745876632,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-8B","ctx":131072,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":0.117,"out":0.455},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.117,"out":0.455,"uptime":99.7,"tps":57,"ttft":640,"reqs":7918}],"scores":{"bfcl":{"v":0.4257,"as":"Qwen3-8B (FC)","effort":"","version":"V4","agent":"BFCL V4 · native function calling","boardAsOf":"2026-04-12"}},"perf":{"uptime":99.7,"n":1,"tps":57,"ttft":640,"via":"Alibaba","reqs":7918},"open":true,"rankable":true,"scored":1},{"id":"recraft/recraft-v3","name":"Recraft: Recraft V3","org":"Recraft","slug":"recraft-v3","matchKey":"recraftv3","blurb":"Recraft V3 is an image generation model from Recraft.","created":1778185433,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000958083832335329},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"ttft":3226,"reqs":3}],"scores":{"arenaImageGen":{"v":1021.4137839893147,"ci":3.7,"rank":68,"votes":191836,"as":"recraft-v3","effort":""}},"perf":{"n":1,"ttft":3226,"via":"Recraft","reqs":3},"open":false,"rankable":true,"scored":1},{"id":"recraft/recraft-v4","name":"Recraft: Recraft V4","org":"Recraft","slug":"recraft-v4","matchKey":"recraftv4","blurb":"Recraft V4 is an image generation model from Recraft.","created":1778185437,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000958083832335329},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100,"ttft":11082,"reqs":10}],"scores":{"arenaImageGen":{"v":1114.2663231329736,"ci":3.4,"rank":47,"votes":114295,"as":"recraft-v4","effort":""}},"perf":{"uptime":100,"n":1,"ttft":11082,"via":"Recraft","reqs":10},"open":false,"rankable":true,"scored":1},{"id":"recraft/recraft-v4.1-pro","name":"Recraft: Recraft V4.1 Pro","org":"Recraft","slug":"recraft-v4.1-pro","matchKey":"recraftv41pro","blurb":"Recraft V4.1 Pro is an image generation model from Recraft tuned for high aesthetics.","created":1778707384,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000502994011976048},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100,"ttft":7803,"reqs":1}],"scores":{"arenaImageGen":{"v":1130.394139424757,"ci":10.7,"rank":39,"votes":2680,"as":"recraft-v4.1-pro","effort":""}},"perf":{"uptime":100,"n":1,"ttft":7803,"via":"Recraft","reqs":1},"open":false,"rankable":true,"scored":1},{"id":"recraft/recraft-v4.1-utility-pro","name":"Recraft: Recraft V4.1 Utility Pro","org":"Recraft","slug":"recraft-v4.1-utility-pro","matchKey":"recraftv41utilitypro","blurb":"Recraft V4.1 Utility Pro is a general-purpose image generation model from Recraft.","created":1778707389,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000502994011976048},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"ttft":13541,"reqs":1}],"scores":{"arenaImageGen":{"v":1169.3842427273012,"ci":11.1,"rank":23,"votes":2516,"as":"recraft-v4.1-utility-pro","effort":""}},"perf":{"n":1,"ttft":13541,"via":"Recraft","reqs":1},"open":false,"rankable":true,"scored":1},{"id":"stepfun/step-3.5-flash","name":"StepFun: Step 3.5 Flash","org":"StepFun","slug":"step-3.5-flash","matchKey":"step35flash","blurb":"Step 3.5 Flash is StepFun's most capable open-source foundation model.","created":1769728337,"cutoff":null,"hf":"stepfun-ai/Step-3.5-Flash","ctx":262144,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.1,"out":0.3},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"SiliconFlow","quant":"fp8","in":0.1,"out":0.3,"uptime":100,"tps":59,"ttft":1966,"reqs":199}],"scores":{"arenaChat":{"v":1394.1342975606894,"ci":3.6,"rank":158,"votes":57137,"as":"step-3.5-flash","effort":""}},"perf":{"uptime":100,"n":1,"tps":59,"ttft":1966,"via":"SiliconFlow","reqs":199},"open":true,"rankable":true,"scored":1},{"id":"tencent/hy4-preview","name":"Tencent: Hy4 preview","org":"Tencent","slug":"hy4-preview","matchKey":"hy4preview","blurb":"Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total.","created":1787897375,"cutoff":null,"hf":"tencent/Hy4-preview","ctx":1048576,"maxOut":64000,"in":["text"],"out":["text"],"price":{"in":0.834,"out":2.501},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","low","none"]},"moderated":false,"providers":[{"name":"Tencent","quant":"fp8","in":0.834,"out":2.501,"cacheRead":0.042,"uptime":100,"tps":41,"ttft":4435,"reqs":553544}],"scores":{"arenaWebdev":{"v":1624.3046280147496,"ci":13.4,"rank":11,"votes":2321,"as":"hy4-preview","effort":""}},"perf":{"uptime":100,"cache":0.042,"n":1,"tps":41,"ttft":4435,"via":"Tencent","reqs":553544},"open":true,"rankable":true,"scored":1},{"id":"upstage/solar-pro-3","name":"Upstage: Solar Pro 3","org":"Upstage","slug":"solar-pro-3","matchKey":"solarpro3","blurb":"Solar Pro 3 is Upstage's powerful Mixture-of-Experts (MoE) language model.","created":1769481200,"cutoff":null,"hf":null,"ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.15,"out":0.6},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","minimal","none"]},"moderated":false,"providers":[{"name":"Upstage","in":0.15,"out":0.6,"cacheRead":0.015,"uptime":100,"tps":125,"ttft":503,"reqs":43}],"scores":{"eqLongform":{"v":27,"as":"solar-pro-3","version":"Longform Writing v1.11"}},"perf":{"uptime":100,"cache":0.015,"n":1,"tps":125,"ttft":503,"via":"Upstage","reqs":43},"open":false,"rankable":true,"scored":1},{"id":"x-ai/grok-imagine-image-quality","name":"SpaceXAI: Grok Imagine Image Quality","org":"SpaceXAI","slug":"grok-imagine-image-quality","matchKey":"grokimagineimagequality","blurb":"Grok Imagine Image Quality is SpaceXAI's fast, high-fidelity image generation and editing model.","created":1779117584,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000119760479041916,"imageIn":0.01},"caps":{"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"xAI","in":0,"out":0,"uptime":100,"ttft":9441,"reqs":218}],"scores":{"arenaImageEdit":{"v":1389.8281145925653,"ci":5.6,"rank":10,"votes":35594,"as":"grok-imagine-image-quality (20260519)","effort":""}},"perf":{"uptime":100,"n":1,"ttft":9441,"via":"xAI","reqs":218},"open":false,"rankable":true,"scored":1},{"id":"xiaomi/mimo-v2.6-pro","name":"Xiaomi: MiMo-V2.6-Pro","org":"Xiaomi","slug":"mimo-v2.6-pro","matchKey":"mimov26pro","blurb":"MiMo-V2.6-Pro is the flagship foundation model developed by Xiaomi.","created":1790021259,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":131072,"in":["text","image","video","audio"],"out":["text"],"price":{"in":0.435,"out":0.87},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Xiaomi","quant":"fp8","in":0.435,"out":0.87,"cacheRead":0.0036,"uptime":100,"tps":44,"ttft":2064,"reqs":20386}],"scores":{"arenaWebdev":{"v":1628.1488731223217,"ci":17.7,"rank":0,"votes":790,"as":"mimo-v2.6-pro","effort":""}},"perf":{"uptime":100,"cache":0.0036,"n":1,"tps":44,"ttft":2064,"via":"Xiaomi","reqs":20386},"open":false,"rankable":true,"scored":1},{"id":"deepseek/deepseek-v4-flash-vision-exp","name":"DeepSeek: DeepSeek V4 Flash Vision Exp","org":"DeepSeek","slug":"deepseek-v4-flash-vision-exp","matchKey":"deepseekv4flashvisionexp","blurb":"DeepSeek V4 Flash Vision Exp is an experimental vision-enabled version of [DeepSeek V4 Flash 0731](https://openrouter.ai/deepseek/deepseek-v4-flash-0731) from DeepSeek, adding image understanding while matching the base ","created":1787311563,"cutoff":null,"hf":"deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","ctx":1048576,"maxOut":943718,"in":["text","image"],"out":["text"],"price":{"in":0.22,"out":0.66,"alt":{"batch":{"in":0.11,"out":0.33}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","maxOut":262144,"in":0.2156,"out":0.6468,"cacheRead":0.00686,"uptime":99.5},{"name":"Fireworks","in":0.22,"out":0.66,"cacheRead":0.007,"uptime":99.9},{"name":"GMICloud","quant":"fp8","ctx":1048575,"maxOut":943717,"in":0.44,"out":1.32,"cacheRead":0.014,"uptime":99.8,"noSchema":1},{"name":"SiliconFlow","quant":"fp8","maxOut":393216,"in":0.44,"out":1.32,"cacheRead":0.028,"uptime":100,"noSchema":1},{"name":"Novita","maxOut":393216,"in":0.44,"out":1.32,"cacheRead":0.028,"uptime":99.9,"noSchema":1}],"scores":{},"perf":{"uptime":100,"cache":0.00686,"n":5},"open":true,"rankable":true,"scored":0},{"id":"openai/gpt-5.6-luna-pro","name":"OpenAI: GPT-5.6 Luna Pro","org":"OpenAI","slug":"gpt-5.6-luna-pro","matchKey":"gpt56lunapro","blurb":"GPT-5.6 Luna Pro is the same underlying model as [GPT-5.6 Luna](https://openrouter.ai/openai/gpt-5.6-luna), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks.","created":1783590867,"cutoff":"2026-02-16","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":0.2,"out":1.2,"webSearch":0.01,"alt":{"batch":{"in":0.1,"out":0.6}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":0.1,"out":0.6,"cacheRead":0.01,"cacheWrite":0.125,"uptime":99.9},{"name":"Azure","in":0.2,"out":1.2,"cacheRead":0.02,"cacheWrite":0.25,"uptime":92.3},{"name":"OpenAI","in":0.2,"out":1.2,"cacheRead":0.02,"cacheWrite":0.25,"uptime":100},{"name":"Azure","region":"eu","in":0.22,"out":1.32,"cacheRead":0.022,"cacheWrite":0.275,"uptime":97},{"name":"OpenAI","in":0.4,"out":2.4,"cacheRead":0.04,"cacheWrite":0.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.01,"regions":["eu"],"n":5},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-5.6-sol-pro","name":"OpenAI: GPT-5.6 Sol Pro","org":"OpenAI","slug":"gpt-5.6-sol-pro","matchKey":"gpt56solpro","blurb":"GPT-5.6 Sol Pro is the same underlying model as [GPT-5.6 Sol](https://openrouter.ai/openai/gpt-5.6-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks.","created":1783590854,"cutoff":"2026-02-16","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":2,"out":10,"webSearch":0.01,"alt":{"batch":{"in":1,"out":5}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":1,"out":5,"cacheRead":0.1,"cacheWrite":1.25,"uptime":69.9},{"name":"OpenAI","in":2,"out":10,"cacheRead":0.2,"cacheWrite":2.5,"uptime":99.8},{"name":"OpenAI","in":4,"out":20,"cacheRead":0.4,"cacheWrite":5,"uptime":100},{"name":"Azure","in":5,"out":30,"cacheRead":0.5,"cacheWrite":6.25,"uptime":100},{"name":"Azure","region":"eu","in":5.5,"out":33,"cacheRead":0.55,"cacheWrite":6.875,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.1,"regions":["eu"],"n":5},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-5.6-terra-pro","name":"OpenAI: GPT-5.6 Terra Pro","org":"OpenAI","slug":"gpt-5.6-terra-pro","matchKey":"gpt56terrapro","blurb":"GPT-5.6 Terra Pro is the same underlying model as [GPT-5.6 Terra](https://openrouter.ai/openai/gpt-5.6-terra), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks.","created":1783590861,"cutoff":"2026-02-16","hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":2,"out":12,"webSearch":0.01,"alt":{"batch":{"in":1,"out":6}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["max","xhigh","high","medium","low","none"]},"moderated":false,"providers":[{"name":"OpenAI","in":1,"out":6,"cacheRead":0.1,"cacheWrite":1.25,"uptime":100},{"name":"Azure","in":2,"out":12,"cacheRead":0.2,"cacheWrite":2.5,"uptime":98},{"name":"OpenAI","in":2,"out":12,"cacheRead":0.2,"cacheWrite":2.5,"uptime":100},{"name":"Azure","region":"eu","in":2.2,"out":13.2,"cacheRead":0.22,"cacheWrite":2.75,"uptime":100},{"name":"OpenAI","in":4,"out":24,"cacheRead":0.4,"cacheWrite":5,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.1,"regions":["eu"],"n":5},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-6-astra-pro","name":"OpenAI: GPT-6 Astra Pro","org":"OpenAI","slug":"gpt-6-astra-pro","matchKey":"gpt6astrapro","blurb":"GPT-6 Astra Pro is the same underlying model as [GPT-6 Astra](https://openrouter.ai/openai/gpt-6-astra), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks.","created":1788552835,"cutoff":null,"hf":null,"ctx":1050000,"maxOut":128000,"in":["file","image","text"],"out":["text"],"price":{"in":10,"out":50,"webSearch":0.01,"alt":{"batch":{"in":5,"out":25}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high","medium","low"]},"moderated":true,"providers":[{"name":"OpenAI","in":5,"out":25,"cacheRead":0.5,"cacheWrite":6.25,"uptime":99.8},{"name":"Azure","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":99.4},{"name":"OpenAI","in":10,"out":50,"cacheRead":1,"cacheWrite":12.5,"uptime":99.9},{"name":"Azure","region":"us","in":11,"out":55,"cacheRead":1.1,"cacheWrite":13.75},{"name":"OpenAI","in":20,"out":100,"cacheRead":2,"cacheWrite":25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.5,"regions":["us"],"n":5},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3-coder","name":"Qwen: Qwen3 Coder 480B A35B","org":"Qwen","slug":"qwen3-coder","matchKey":"qwen3coder","blurb":"Qwen3-Coder-480B-A35B-Instruct is a Mixture-of-Experts (MoE) code generation model developed by the Qwen team.","created":1753230546,"cutoff":"2025-06-30","hf":"Qwen/Qwen3-Coder-480B-A35B-Instruct","ctx":262144,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.3,"out":1},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google","region":"us","in":0.22,"out":1.8,"uptime":100},{"name":"DeepInfra","quant":"fp4","in":0.3,"out":1,"cacheRead":0.1,"uptime":96.2},{"name":"Venice","quant":"fp8","ctx":256000,"in":0.35,"out":1.5,"cacheRead":0.04,"uptime":87},{"name":"Novita","quant":"fp8","in":0.38,"out":1.55,"uptime":94.5},{"name":"Alibaba","in":0.975,"out":4.875,"uptime":99.9}],"scores":{},"perf":{"uptime":100,"regions":["us"],"n":5},"open":true,"rankable":true,"scored":0},{"id":"nvidia/nemotron-3-nano-30b-a3b","name":"NVIDIA: Nemotron 3 Nano 30B A3B","org":"NVIDIA","slug":"nemotron-3-nano-30b-a3b","matchKey":"nemotron3nano30ba3b","blurb":"NVIDIA Nemotron 3 Nano 30B A3B is a small language MoE model with highest compute efficiency and accuracy for developers to build specialized agentic AI systems.","created":1765731275,"cutoff":null,"hf":"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.05,"out":0.2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Crusoe","quant":"fp8","in":0.05,"out":0.2,"cacheRead":0.03,"uptime":100},{"name":"Novita","quant":"fp4","maxOut":32768,"in":0.05,"out":0.2,"uptime":100,"noSchema":1},{"name":"DeepInfra","quant":"fp4","maxOut":228000,"in":0.05,"out":0.2,"cacheRead":0.025,"uptime":99,"noSchema":1},{"name":"Nebius","quant":"fp8","in":0.06,"out":0.24,"uptime":99.8}],"scores":{},"perf":{"uptime":100,"cache":0.03,"n":4},"open":true,"rankable":true,"scored":0},{"id":"nvidia/nemotron-3-ultra-550b-a55b","name":"NVIDIA: Nemotron 3 Ultra","org":"NVIDIA","slug":"nemotron-3-ultra-550b-a55b","matchKey":"nemotron3ultra550ba55b","blurb":"NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE).","created":1780551208,"cutoff":null,"hf":"nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16","ctx":262144,"maxOut":182520,"in":["text"],"out":["text"],"price":{"in":0.6,"out":2.4,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","medium"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","maxOut":16384,"in":0.5,"out":2.2,"cacheRead":0.1,"uptime":99.8},{"name":"BaseTen","quant":"fp4","ctx":202800,"in":0.6,"out":2.4,"cacheRead":0.12,"uptime":100,"noSchema":1},{"name":"BaseTen","quant":"fp4","ctx":202800,"in":0.6,"out":2.4,"cacheRead":0.12,"uptime":100,"noSchema":1},{"name":"Venice","quant":"fp8","ctx":256000,"maxOut":32768,"in":0.625,"out":3.125,"cacheRead":0.1875,"uptime":90.1,"noSchema":1}],"scores":{},"perf":{"uptime":100,"cache":0.1,"n":4},"open":true,"rankable":true,"scored":0},{"id":"nvidia/nemotron-3.5-lightning","name":"NVIDIA: Nemotron 3.5 Lightning","org":"NVIDIA","slug":"nemotron-3.5-lightning","matchKey":"nemotron35lightning","blurb":"NVIDIA Nemotron 3.5 Lightning is an open mixture-of-experts model from NVIDIA, with 3B active parameters out of 30B total.","created":1786452751,"cutoff":null,"hf":"nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.07,"out":0.2,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"int4","maxOut":32768,"in":0.065,"out":0.18,"uptime":100,"noTools":1},{"name":"Phala","in":0.07,"out":0.2,"cacheRead":0.04,"uptime":99.2},{"name":"CoreWeave","quant":"bf16","in":0.07,"out":0.2,"cacheRead":0.04,"uptime":100},{"name":"DeepInfra","quant":"bf16","maxOut":131072,"in":0.08,"out":0.2,"cacheRead":0.04,"uptime":99.9}],"scores":{},"perf":{"uptime":100,"n":4},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-coder-next","name":"Qwen: Qwen3 Coder Next","org":"Qwen","slug":"qwen3-coder-next","matchKey":"qwen3codernext","blurb":"Qwen3-Coder-Next is an open-weight causal language model optimized for coding agents and local development workflows.","created":1770164101,"cutoff":null,"hf":"Qwen/Qwen3-Coder-Next","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.12,"out":0.8},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"bf16","in":0.12,"out":0.8,"cacheRead":0.07,"uptime":100},{"name":"StreamLake","ctx":256000,"maxOut":64000,"in":0.18,"out":0.9,"cacheRead":0.036,"uptime":99.3,"noSchema":1},{"name":"Novita","quant":"fp8","maxOut":65536,"in":0.2,"out":1.5,"uptime":99.7,"noSchema":1},{"name":"Alibaba","maxOut":65536,"in":0.3,"out":1.5,"uptime":99.1,"noSchema":1}],"scores":{},"perf":{"uptime":100,"cache":0.07,"n":4},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-vl-30b-a3b-instruct","name":"Qwen: Qwen3 VL 30B A3B Instruct","org":"Qwen","slug":"qwen3-vl-30b-a3b-instruct","matchKey":"qwen3vl30ba3binstruct","blurb":"Qwen3-VL-30B-A3B-Instruct is a multimodal model that unifies strong text generation with visual understanding for images and videos.","created":1759794476,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-VL-30B-A3B-Instruct","ctx":262144,"maxOut":32768,"in":["text","image"],"out":["text"],"price":{"in":0.13,"out":0.52},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","ctx":131072,"in":0.13,"out":0.52,"uptime":99.9},{"name":"DeepInfra","quant":"fp8","maxOut":16384,"in":0.15,"out":0.6,"uptime":99.2},{"name":"Novita","quant":"bf16","ctx":131072,"in":0.2,"out":0.7,"uptime":97.7},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.29,"out":1,"uptime":94.4}],"scores":{},"perf":{"uptime":99.9,"n":4},"open":true,"rankable":true,"scored":0},{"id":"x-ai/grok-4.20","name":"SpaceXAI: Grok 4.20","org":"SpaceXAI","slug":"grok-4.20","matchKey":"grok420","blurb":"Grok 4.20 is a reasoning model from SpaceXAI with industry-leading speed and agentic tool calling capabilities.","created":1774979019,"cutoff":"2025-09-01","hf":null,"ctx":2000000,"maxOut":1800000,"in":["text","image","file"],"out":["text"],"price":{"in":1.25,"out":2.5,"webSearch":0.005},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"xAI","in":1.25,"out":2.5,"cacheRead":0.2,"uptime":99.6},{"name":"xAI","in":1.25,"out":2.5,"cacheRead":0.2,"uptime":99.7},{"name":"xAI","in":2.5,"out":5,"cacheRead":0.4,"uptime":100},{"name":"xAI","in":2.5,"out":5,"cacheRead":0.4,"uptime":99.5}],"scores":{},"perf":{"uptime":100,"cache":0.2,"n":4},"open":false,"rankable":true,"scored":0},{"id":"x-ai/grok-build-0.1","name":"SpaceXAI: Grok Build 0.1","org":"SpaceXAI","slug":"grok-build-0.1","matchKey":"grokbuild01","blurb":"Grok Build 0.1 is SpaceXAI’s fast coding model trained specifically for agentic software engineering workflows.","created":1779298123,"cutoff":null,"hf":null,"ctx":256000,"maxOut":230400,"in":["text","image","file"],"out":["text"],"price":{"in":1,"out":2,"webSearch":0.005},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"xAI","in":1,"out":2,"cacheRead":0.2,"uptime":99.9},{"name":"xAI","in":1,"out":2,"cacheRead":0.2,"uptime":100},{"name":"xAI","in":2,"out":4,"cacheRead":0.4},{"name":"xAI","in":2,"out":4,"cacheRead":0.4}],"scores":{},"perf":{"uptime":100,"cache":0.2,"n":4},"open":false,"rankable":true,"scored":0},{"id":"gryphe/mythomax-l2-13b","name":"MythoMax 13B","org":"gryphe","slug":"mythomax-l2-13b","matchKey":"mythomaxl213b","blurb":"One of the highest performing and most popular fine-tunes of Llama 2 13B, with rich descriptions and roleplay.","created":1688256000,"cutoff":"2023-06-30","hf":"Gryphe/MythoMax-L2-13b","ctx":8192,"maxOut":3686,"in":["text"],"out":["text"],"price":{"in":0.08,"out":0.11},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"fp16","ctx":4096,"in":0.08,"out":0.11,"uptime":100},{"name":"Mancer 2","quant":"fp8","maxOut":7372,"in":0.35,"out":0.6,"uptime":100},{"name":"DeepInfra","quant":"fp16","ctx":4096,"in":0.4,"out":0.4,"uptime":100,"noSchema":1}],"scores":{},"perf":{"uptime":100,"n":3},"open":true,"rankable":true,"scored":0},{"id":"mistralai/ministral-3b-2512","name":"Mistral: Ministral 3 3B 2512","org":"Mistral","slug":"ministral-3b-2512","matchKey":"ministral3b2512","blurb":"The smallest model in the Ministral 3 family, Ministral 3 3B is a powerful, efficient tiny language model with vision capabilities.","created":1764681560,"cutoff":null,"hf":"mistralai/Ministral-3-3B-Instruct-2512","ctx":131072,"maxOut":104857,"in":["text","image"],"out":["text"],"price":{"in":0.1,"out":0.1},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.1,"out":0.1,"cacheRead":0.01,"uptime":99.6},{"name":"Mistral","in":0.1,"out":0.1,"cacheRead":0.01,"uptime":99.8},{"name":"Mistral","region":"eu","in":0.11,"out":0.11,"cacheRead":0.011,"uptime":99.9}],"scores":{},"perf":{"uptime":99.9,"cache":0.01,"regions":["eu"],"n":3},"open":true,"rankable":true,"scored":0},{"id":"mistralai/ministral-8b-2512","name":"Mistral: Ministral 3 8B 2512","org":"Mistral","slug":"ministral-8b-2512","matchKey":"ministral8b2512","blurb":"A balanced model in the Ministral 3 family, Ministral 3 8B is a powerful, efficient tiny language model with vision capabilities.","created":1764681654,"cutoff":null,"hf":"mistralai/Ministral-3-8B-Instruct-2512","ctx":262144,"maxOut":209715,"in":["text","image"],"out":["text"],"price":{"in":0.15,"out":0.15,"alt":{"batch":{"in":0.075,"out":0.075}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.15,"out":0.15,"cacheRead":0.015,"uptime":98.6},{"name":"Mistral","in":0.15,"out":0.15,"cacheRead":0.015,"uptime":98.4},{"name":"Mistral","region":"eu","in":0.165,"out":0.165,"cacheRead":0.0165,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.015,"regions":["eu"],"n":3},"open":true,"rankable":true,"scored":0},{"id":"mistralai/mistral-large-2407","name":"Mistral Large 2407","org":"mistralai","slug":"mistral-large-2407","matchKey":"mistrallarge2407","blurb":"This is Mistral AI's flagship model, Mistral Large 2 (version mistral-large-2407).","created":1731978415,"cutoff":"2024-03-31","hf":null,"ctx":131072,"maxOut":104857,"in":["text","file"],"out":["text"],"price":{"in":2,"out":6},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":2,"out":6,"cacheRead":0.2,"uptime":99.6},{"name":"Mistral","in":2,"out":6,"cacheRead":0.2,"uptime":99.6},{"name":"Mistral","region":"eu","in":2.2,"out":6.6,"cacheRead":0.22}],"scores":{},"perf":{"uptime":99.6,"cache":0.2,"regions":["eu"],"n":3},"open":false,"rankable":true,"scored":0},{"id":"mistralai/mistral-saba","name":"Mistral: Saba","org":"Mistral","slug":"mistral-saba","matchKey":"mistralsaba","blurb":"Mistral Saba is a 24B-parameter language model specifically designed for the Middle East and South Asia, delivering accurate and contextually relevant responses while maintaining efficient performance.","created":1739803239,"cutoff":"2024-09-30","hf":null,"ctx":32768,"maxOut":26214,"in":["text","file"],"out":["text"],"price":{"in":0.2,"out":0.6},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.2,"out":0.6,"cacheRead":0.02,"uptime":100},{"name":"Mistral","in":0.2,"out":0.6,"cacheRead":0.02,"uptime":100},{"name":"Mistral","region":"eu","in":0.22,"out":0.66,"cacheRead":0.022}],"scores":{},"perf":{"uptime":100,"cache":0.02,"regions":["eu"],"n":3},"open":false,"rankable":true,"scored":0},{"id":"mistralai/mixtral-8x22b-instruct","name":"Mistral: Mixtral 8x22B Instruct","org":"Mistral","slug":"mixtral-8x22b-instruct","matchKey":"mixtral8x22binstruct","blurb":"Mistral's official instruct fine-tuned version of [Mixtral 8x22B](/models/mistralai/mixtral-8x22b).","created":1713312000,"cutoff":"2024-01-31","hf":"mistralai/Mixtral-8x22B-Instruct-v0.1","ctx":65536,"maxOut":52428,"in":["text","file"],"out":["text"],"price":{"in":2,"out":6},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":2,"out":6,"cacheRead":0.2,"uptime":100},{"name":"Mistral","in":2,"out":6,"cacheRead":0.2,"uptime":100},{"name":"Mistral","region":"eu","in":2.2,"out":6.6,"cacheRead":0.22,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.2,"regions":["eu"],"n":3},"open":true,"rankable":true,"scored":0},{"id":"mistralai/voxtral-small-24b-2507","name":"Mistral: Voxtral Small 24B 2507","org":"Mistral","slug":"voxtral-small-24b-2507","matchKey":"voxtralsmall24b2507","blurb":"Voxtral Small is an enhancement of Mistral Small 3, incorporating state-of-the-art audio input capabilities while retaining best-in-class text performance.","created":1761835144,"cutoff":null,"hf":"mistralai/Voxtral-Small-24B-2507","ctx":32768,"maxOut":26214,"in":["text","audio","file"],"out":["text"],"price":{"in":0.1,"out":0.3,"audioIn":100},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.1,"out":0.3,"cacheRead":0.01,"uptime":99.7},{"name":"Mistral","ctx":32000,"maxOut":25600,"in":0.1,"out":0.3,"cacheRead":0.01,"uptime":99.8},{"name":"Mistral","region":"eu","ctx":32000,"maxOut":25600,"in":0.11,"out":0.33,"cacheRead":0.011}],"scores":{},"perf":{"uptime":99.8,"cache":0.01,"regions":["eu"],"n":3},"open":true,"rankable":true,"scored":0},{"id":"openai/whisper-large-v3","name":"OpenAI: Whisper Large V3","org":"OpenAI","slug":"whisper-large-v3","matchKey":"whisperlargev3","blurb":"Whisper Large V3 is OpenAI's open-source automatic speech recognition model offering both audio transcription and translation.","created":1777642266,"cutoff":null,"hf":"openai/whisper-large-v3","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":7.5,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","in":7.5,"out":0,"uptime":100},{"name":"Together","in":25,"out":0,"uptime":100},{"name":"Groq","in":30.833333,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":3},"open":true,"rankable":true,"scored":0},{"id":"sao10k/l3-lunaris-8b","name":"Sao10K: Llama 3 8B Lunaris","org":"Sao10K","slug":"l3-lunaris-8b","matchKey":"l3lunaris8b","blurb":"Lunaris 8B is a versatile generalist and roleplaying model based on Llama 3.","created":1723507200,"cutoff":"2023-12-31","hf":"Sao10K/L3-8B-Lunaris-v1","ctx":8192,"maxOut":7372,"in":["text"],"out":["text"],"price":{"in":0.04,"out":0.05},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"bf16","in":0.04,"out":0.05,"uptime":100},{"name":"DeepInfra","quant":"fp8","in":0.04,"out":0.05,"uptime":100,"noSchema":1},{"name":"Novita","quant":"bf16","in":0.05,"out":0.05,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":3},"open":true,"rankable":true,"scored":0},{"id":"stepfun/step-3.7-flash","name":"StepFun: Step 3.7 Flash","org":"StepFun","slug":"step-3.7-flash","matchKey":"step37flash","blurb":"Step 3.7 Flash is StepFun's latest high-efficiency multimodal Mixture-of-Experts model.","created":1779985069,"cutoff":null,"hf":"stepfun-ai/Step-3.7-Flash","ctx":262144,"maxOut":230400,"in":["text","image","video"],"out":["text"],"price":{"in":0.2,"out":1.15},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"DeepInfra","maxOut":32768,"in":0.16,"out":0.92,"cacheRead":0.032,"uptime":99.9,"noSchema":1},{"name":"Novita","quant":"fp8","maxOut":256000,"in":0.2,"out":1.15,"cacheRead":0.04,"uptime":99.3},{"name":"StepFun","quant":"fp8","ctx":256000,"in":0.2,"out":1.15,"cacheRead":0.04,"uptime":99.3}],"scores":{},"perf":{"uptime":99.9,"cache":0.032,"n":3},"open":true,"rankable":true,"scored":0},{"id":"amazon/nova-lite-v1","name":"Amazon: Nova Lite 1.0","org":"Amazon","slug":"nova-lite-v1","matchKey":"novalitev1","blurb":"Amazon Nova Lite 1.0 is a very low-cost multimodal model from Amazon that focused on fast processing of image, video, and text inputs to generate text output.","created":1733437363,"cutoff":"2024-10-31","hf":null,"ctx":300000,"maxOut":5120,"in":["text","image"],"out":["text"],"price":{"in":0.06,"out":0.24},"caps":{"tools":true},"reasoning":null,"moderated":true,"providers":[{"name":"Amazon Bedrock","region":"eu","in":0.06,"out":0.24,"uptime":89.7},{"name":"Amazon Bedrock","in":0.06,"out":0.24,"uptime":100}],"scores":{},"perf":{"uptime":100,"regions":["eu"],"n":2},"open":false,"rankable":true,"scored":0},{"id":"deepseek/deepseek-chat","name":"DeepSeek: DeepSeek V3","org":"DeepSeek","slug":"deepseek-chat","matchKey":"deepseekchat","blurb":"DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following and coding abilities of the previous versions.","created":1735241320,"cutoff":"2024-07-31","hf":"deepseek-ai/DeepSeek-V3","ctx":163840,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.32,"out":0.89},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"StreamLake","ctx":128000,"maxOut":16000,"in":0.2574,"out":1.0287,"uptime":94.8},{"name":"DeepInfra","quant":"fp4","in":0.32,"out":0.89,"uptime":97}],"scores":{},"perf":{"uptime":97,"n":2},"open":true,"rankable":true,"scored":0},{"id":"ibm-granite/granite-4.2-8b","name":"IBM: Granite 4.2 8B","org":"IBM","slug":"granite-4.2-8b","matchKey":"granite428b","blurb":"Granite 4.2 8B is a dense reasoning model from IBM.","created":1788206780,"cutoff":null,"hf":"ibm-granite/granite-4.2-8b","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.06,"out":0.25},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","low","none"]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":0.06,"out":0.25,"cacheRead":0.015,"uptime":100,"noSchema":1},{"name":"CoreWeave","quant":"bf16","in":0.1,"out":0.15,"cacheRead":0.05,"uptime":99.7}],"scores":{},"perf":{"uptime":100,"cache":0.015,"n":2},"open":true,"rankable":true,"scored":0},{"id":"inclusionai/ling-3.0-flash","name":"inclusionAI: Ling 3.0 Flash","org":"inclusionAI","slug":"ling-3.0-flash","matchKey":"ling30flash","blurb":"*Ling-3.0-flash* is a *124B-parameter Mixture-of-Experts (MoE) model*, with approximately *5.1B parameters activated per token*.","created":1784818580,"cutoff":null,"hf":"inclusionAI/Ling-3.0-flash","ctx":262144,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.021,"out":0.063},"caps":{"tools":true,"json":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","in":0.021,"out":0.063,"cacheRead":0.0042,"uptime":100},{"name":"DeepInfra","quant":"bf16","ctx":131072,"in":0.06,"out":0.18,"cacheRead":0.012,"uptime":99.6}],"scores":{},"perf":{"uptime":100,"cache":0.0042,"n":2},"open":true,"rankable":true,"scored":0},{"id":"mistralai/codestral-2508","name":"Mistral: Codestral 2508","org":"Mistral","slug":"codestral-2508","matchKey":"codestral2508","blurb":"Mistral's cutting-edge language model for coding released end of July 2025.","created":1754079630,"cutoff":"2025-03-31","hf":null,"ctx":256000,"maxOut":204800,"in":["text","file"],"out":["text"],"price":{"in":0.3,"out":0.9,"alt":{"batch":{"in":0.15,"out":0.45}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":0.3,"out":0.9,"cacheRead":0.03,"uptime":99.8},{"name":"Mistral","region":"eu","in":0.33,"out":0.99,"cacheRead":0.033}],"scores":{},"perf":{"uptime":99.8,"cache":0.03,"regions":["eu"],"n":2},"open":false,"rankable":true,"scored":0},{"id":"mistralai/mistral-large","name":"Mistral Large","org":"mistralai","slug":"mistral-large","matchKey":"mistrallarge","blurb":"This is Mistral AI's flagship model, Mistral Large 2 (version `mistral-large-2407`).","created":1708905600,"cutoff":"2024-11-30","hf":null,"ctx":128000,"maxOut":102400,"in":["text","file"],"out":["text"],"price":{"in":2,"out":6},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":2,"out":6,"cacheRead":0.2,"uptime":99.8},{"name":"Mistral","region":"eu","in":2.2,"out":6.6,"cacheRead":0.22,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.2,"regions":["eu"],"n":2},"open":false,"rankable":true,"scored":0},{"id":"mistralai/voxtral-mini-transcribe","name":"Mistral: Voxtral Mini Transcribe","org":"Mistral","slug":"voxtral-mini-transcribe","matchKey":"voxtralminitranscribe","blurb":"Voxtral Mini Transcribe is Mistral's speech-to-text model, derived from the Voxtral Mini family.","created":1778877024,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":50,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Mistral","in":50,"out":0,"uptime":100},{"name":"Mistral","region":"eu","in":55,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"regions":["eu"],"n":2},"open":false,"rankable":true,"scored":0},{"id":"nvidia/nemotron-3-super-120b-a12b","name":"NVIDIA: Nemotron 3 Super","org":"NVIDIA","slug":"nemotron-3-super-120b-a12b","matchKey":"nemotron3super120ba12b","blurb":"NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications.","created":1773245239,"cutoff":null,"hf":"nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.08,"out":0.45,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["medium","low"]},"moderated":false,"providers":[{"name":"DekaLLM","quant":"fp8","in":0.08,"out":0.45,"uptime":97.7},{"name":"DeepInfra","quant":"bf16","maxOut":16384,"in":0.085,"out":0.4,"uptime":100,"noSchema":1}],"scores":{},"perf":{"uptime":100,"n":2},"open":true,"rankable":true,"scored":0},{"id":"openai/gpt-3.5-turbo-16k","name":"OpenAI: GPT-3.5 Turbo 16k","org":"OpenAI","slug":"gpt-3.5-turbo-16k","matchKey":"gpt35turbo16k","blurb":"This model offers four times the context length of gpt-3.5-turbo, allowing it to support approximately 20 pages of text in a single request at a higher cost.","created":1693180800,"cutoff":"2021-09-30","hf":null,"ctx":16385,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":3,"out":4},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":3,"out":4,"uptime":100},{"name":"OpenAI","in":3,"out":4,"uptime":99.7}],"scores":{},"perf":{"uptime":100,"n":2},"open":false,"rankable":true,"scored":0},{"id":"openai/whisper-large-v3-turbo","name":"OpenAI: Whisper Large V3 Turbo","org":"OpenAI","slug":"whisper-large-v3-turbo","matchKey":"whisperlargev3turbo","blurb":"Whisper Large V3 Turbo is an optimized version of OpenAI's Whisper Large V3 speech recognition model, designed for speed and cost efficiency.","created":1777642266,"cutoff":null,"hf":"openai/whisper-large-v3-turbo","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":3.33,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","in":3.33,"out":0,"uptime":100},{"name":"Groq","in":11.111111,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":2},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen-2.5-72b-instruct","name":"Qwen2.5 72B Instruct","org":"qwen","slug":"qwen-2.5-72b-instruct","matchKey":"qwen2572binstruct","blurb":"Qwen2.5 72B is the latest series of Qwen large language models.","created":1726704000,"cutoff":"2024-06-30","hf":"Qwen/Qwen2.5-72B-Instruct","ctx":32768,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.36,"out":0.4},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.36,"out":0.4,"uptime":99.7},{"name":"Novita","quant":"bf16","ctx":32000,"maxOut":8192,"in":0.38,"out":0.4,"uptime":97.1,"noSchema":1}],"scores":{},"perf":{"uptime":99.7,"n":2},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-vl-30b-a3b-thinking","name":"Qwen: Qwen3 VL 30B A3B Thinking","org":"Qwen","slug":"qwen3-vl-30b-a3b-thinking","matchKey":"qwen3vl30ba3bthinking","blurb":"Qwen3-VL-30B-A3B-Thinking is a multimodal model that unifies strong text generation with visual understanding for images and videos.","created":1759794479,"cutoff":"2025-03-31","hf":"Qwen/Qwen3-VL-30B-A3B-Thinking","ctx":262144,"maxOut":32768,"in":["text","image"],"out":["text"],"price":{"in":0.2,"out":2.4},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","ctx":131072,"in":0.2,"out":2.4,"uptime":100},{"name":"SiliconFlow","quant":"fp8","maxOut":235929,"in":0.29,"out":1,"uptime":99.7}],"scores":{},"perf":{"uptime":100,"n":2},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-vl-8b-instruct","name":"Qwen: Qwen3 VL 8B Instruct","org":"Qwen","slug":"qwen3-vl-8b-instruct","matchKey":"qwen3vl8binstruct","blurb":"Qwen3-VL-8B-Instruct is a multimodal vision-language model from the Qwen3-VL series, built for high-fidelity understanding and reasoning across text, images, and video.","created":1760463308,"cutoff":null,"hf":"Qwen/Qwen3-VL-8B-Instruct","ctx":262144,"maxOut":32768,"in":["image","text"],"out":["text"],"price":{"in":0.117,"out":0.455},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","ctx":131072,"in":0.117,"out":0.455,"uptime":99.9},{"name":"Parasail","quant":"bf16","maxOut":235929,"in":0.25,"out":0.75,"cacheRead":0.12,"uptime":98.6}],"scores":{},"perf":{"uptime":99.9,"n":2},"open":true,"rankable":true,"scored":0},{"id":"sao10k/l3.1-euryale-70b","name":"Sao10K: Llama 3.1 Euryale 70B v2.2","org":"Sao10K","slug":"l3.1-euryale-70b","matchKey":"l31euryale70b","blurb":"Euryale L3.1 70B v2.2 is a model focused on creative roleplay from [Sao10k](https://ko-fi.com/sao10k).","created":1724803200,"cutoff":"2023-12-31","hf":"Sao10K/L3.1-70B-Euryale-v2.2","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.85,"out":0.85},"caps":{"tools":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.85,"out":0.85,"uptime":99.6,"noTools":1},{"name":"Novita","quant":"fp8","ctx":8192,"maxOut":7372,"in":1.4504,"out":1.4504,"uptime":99.7}],"scores":{},"perf":{"uptime":99.7,"n":2},"open":true,"rankable":true,"scored":0},{"id":"undi95/remm-slerp-l2-13b","name":"ReMM SLERP 13B","org":"undi95","slug":"remm-slerp-l2-13b","matchKey":"remmslerpl213b","blurb":"A recreation trial of the original MythoMax-L2-B13 but with updated models.","created":1689984000,"cutoff":"2023-06-30","hf":"Undi95/ReMM-SLERP-L2-13B","ctx":6144,"maxOut":5529,"in":["text"],"out":["text"],"price":{"in":0.35,"out":0.65},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mancer 2","quant":"fp8","in":0.35,"out":0.65,"uptime":100},{"name":"NextBit","quant":"bf16","maxOut":4096,"in":0.45,"out":0.65,"uptime":98.5}],"scores":{},"perf":{"uptime":100,"n":2},"open":true,"rankable":true,"scored":0},{"id":"aion-labs/aion-2.0","name":"AionLabs: Aion-2.0","org":"AionLabs","slug":"aion-2.0","matchKey":"aion20","blurb":"Aion-2.0 is a variant of DeepSeek V3.2 optimized for immersive roleplaying and storytelling.","created":1771881306,"cutoff":null,"hf":null,"ctx":131072,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.8,"out":1.6},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"AionLabs","in":0.8,"out":1.6,"cacheRead":0.2,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.2,"n":1},"open":false,"rankable":true,"scored":0},{"id":"aion-labs/aion-3.0","name":"AionLabs: Aion-3.0","org":"AionLabs","slug":"aion-3.0","matchKey":"aion30","blurb":"Aion-3.0 is a multi-model roleplaying and storytelling system from AionLabs, built on the GLM family of models.","created":1783443095,"cutoff":null,"hf":null,"ctx":131072,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":3,"out":6},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"AionLabs","in":3,"out":6,"cacheRead":0.75,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.75,"n":1},"open":false,"rankable":true,"scored":0},{"id":"aion-labs/aion-3.0-mini","name":"AionLabs: Aion-3.0-Mini","org":"AionLabs","slug":"aion-3.0-mini","matchKey":"aion30mini","blurb":"Aion-3.0 Mini is a multi-model roleplaying and storytelling system from AionLabs, built on the DeepSeek family of models.","created":1783443096,"cutoff":null,"hf":null,"ctx":131072,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.7,"out":1.4},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"AionLabs","in":0.7,"out":1.4,"cacheRead":0.18,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.18,"n":1},"open":false,"rankable":true,"scored":0},{"id":"aion-labs/aion-rp-llama-3.1-8b","name":"AionLabs: Aion-RP 1.0 (8B)","org":"AionLabs","slug":"aion-rp-llama-3.1-8b","matchKey":"aionrpllama318b","blurb":"Aion-RP-Llama-3.1-8B ranks the highest in the character evaluation portion of the RPBench-Auto benchmark, a roleplaying-specific variant of Arena-Hard-Auto, where LLMs evaluate each other’s responses.","created":1738696718,"cutoff":"2023-12-31","hf":null,"ctx":32768,"maxOut":29491,"in":["text"],"out":["text"],"price":{"in":0.8,"out":1.6},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"AionLabs","in":0.8,"out":1.6,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"alibaba/happyhorse-1.1","name":"Alibaba: HappyHorse 1.1","org":"Alibaba","slug":"happyhorse-1.1","matchKey":"happyhorse11","blurb":"HappyHorse 1.1 is a video generation model from Alibaba.","created":1782269643,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"alibaba/wan-2.6","name":"Alibaba: Wan 2.6","org":"Alibaba","slug":"wan-2.6","matchKey":"wan26","blurb":"Alibaba's most advanced video generation model, supporting over 10 visual creation capabilities in a unified system.","created":1774659190,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"alibaba/wan-2.7","name":"Alibaba: Wan 2.7","org":"Alibaba","slug":"wan-2.7","matchKey":"wan27","blurb":"Wan 2.7 is a video generation model from Alibaba.","created":1776211362,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"alibaba/wan-3.0-prime","name":"Alibaba: Wan 3.0 Prime","org":"Alibaba","slug":"wan-3.0-prime","matchKey":"wan30prime","blurb":"Wan 3.0 Prime is a fast-mode variant of [Wan 3.0](https://openrouter.ai/alibaba/wan-3.0) from Alibaba.","created":1787863800,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"amazon/nova-premier-v1","name":"Amazon: Nova Premier 1.0","org":"Amazon","slug":"nova-premier-v1","matchKey":"novapremierv1","blurb":"Amazon Nova Premier is the most capable of Amazon’s multimodal models for complex reasoning tasks and for use as the best teacher for distilling custom models.","created":1761950332,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":32000,"in":["text","image"],"out":["text"],"price":{"in":2.5,"out":12.5},"caps":{"tools":true},"reasoning":null,"moderated":true,"providers":[{"name":"Amazon Bedrock","in":2.5,"out":12.5,"cacheRead":0.625}],"scores":{},"perf":{"cache":0.625,"n":1},"open":false,"rankable":true,"scored":0},{"id":"anthracite-org/magnum-v4-72b","name":"Magnum v4 72B","org":"anthracite-org","slug":"magnum-v4-72b","matchKey":"magnumv472b","blurb":"This is a series of models designed to replicate the prose quality of the Claude 3 models, specifically Sonnet(https://openrouter.ai/anthropic/claude-3.5-sonnet) and Opus(https://openrouter.ai/anthropic/claude-3-opus).","created":1729555200,"cutoff":"2024-06-30","hf":"anthracite-org/magnum-v4-72b","ctx":32768,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":2.5,"out":5},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mancer 2","quant":"fp8","in":2.5,"out":5,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"baidu/ernie-4.5-vl-424b-a47b","name":"Baidu: ERNIE 4.5 VL 424B A47B ","org":"Baidu","slug":"ernie-4.5-vl-424b-a47b","matchKey":"ernie45vl424ba47b","blurb":"ERNIE-4.5-VL-424B-A47B is a multimodal Mixture-of-Experts (MoE) model from Baidu’s ERNIE 4.5 series, featuring 424B total parameters with 47B active per token.","created":1751300903,"cutoff":"2025-03-31","hf":"baidu/ERNIE-4.5-VL-424B-A47B-PT","ctx":123000,"maxOut":16000,"in":["image","text"],"out":["text"],"price":{"in":0.42,"out":1.25},"caps":{"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"fp16","in":0.42,"out":1.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"black-forest-labs/flux-video-edit","name":"Black Forest Labs: FLUX Video Edit","org":"Black Forest Labs","slug":"flux-video-edit","matchKey":"fluxvideoedit","blurb":"FLUX Video Edit [fast] takes a source video and an edit prompt and returns a precisely edited video.","created":1789073532,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","video"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"black-forest-labs/flux-video-upscale","name":"Black Forest Labs: FLUX Video Upscale","org":"Black Forest Labs","slug":"flux-video-upscale","matchKey":"fluxvideoupscale","blurb":"FLUX Video Upscale is a video upscaling model from Black Forest Labs.","created":1787173919,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","video"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Black Forest Labs","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance-seed/seed-1.6","name":"ByteDance Seed: Seed 1.6","org":"ByteDance Seed","slug":"seed-1.6","matchKey":"seed16","blurb":"Seed 1.6 is a general-purpose model released by the ByteDance Seed team.","created":1766504997,"cutoff":null,"hf":null,"ctx":262144,"maxOut":32768,"in":["image","text","video"],"out":["text"],"price":{"in":0.25,"out":2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Seed","quant":"fp8","in":0.25,"out":2,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance-seed/seed-1.6-flash","name":"ByteDance Seed: Seed 1.6 Flash","org":"ByteDance Seed","slug":"seed-1.6-flash","matchKey":"seed16flash","blurb":"Seed 1.6 Flash is an ultra-fast multimodal deep thinking model by ByteDance Seed, supporting both text and visual understanding.","created":1766505011,"cutoff":null,"hf":null,"ctx":262144,"maxOut":32768,"in":["image","text","video"],"out":["text"],"price":{"in":0.075,"out":0.3},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Seed","quant":"fp8","in":0.075,"out":0.3,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance-seed/seed-2-1-turbo","name":"ByteDance Seed: Seed 2.1 Turbo","org":"ByteDance Seed","slug":"seed-2-1-turbo","matchKey":"seed21turbo","blurb":"Seed 2.1 Turbo is a multimodal model from ByteDance Seed for coding and long-horizon agent workflows.","created":1786552176,"cutoff":null,"hf":null,"ctx":262144,"maxOut":235929,"in":["text","image","video"],"out":["text"],"price":{"in":0.5,"out":2.5},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Seed","quant":"fp8","in":0.5,"out":2.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance-seed/seed-2.0-code","name":"ByteDance Seed: Seed-2.0-Code","org":"ByteDance Seed","slug":"seed-2.0-code","matchKey":"seed20code","blurb":"Seed 2.0 Code is a model from ByteDance Seed optimized for agentic coding.","created":1786550701,"cutoff":null,"hf":null,"ctx":262144,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":0.5,"out":3},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Seed","quant":"fp8","in":0.5,"out":3,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance-seed/seed-2.0-lite","name":"ByteDance Seed: Seed-2.0-Lite","org":"ByteDance Seed","slug":"seed-2.0-lite","matchKey":"seed20lite","blurb":"Seed-2.0-Lite is a versatile, cost‑efficient enterprise workhorse that delivers strong multimodal and agent capabilities while offering noticeably lower latency, making it a practical default choice for most production w","created":1773157231,"cutoff":null,"hf":null,"ctx":262144,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":0.25,"out":2},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Seed","quant":"fp8","in":0.25,"out":2,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance-seed/seed-2.0-mini","name":"ByteDance Seed: Seed-2.0-Mini","org":"ByteDance Seed","slug":"seed-2.0-mini","matchKey":"seed20mini","blurb":"Seed-2.0-mini targets latency-sensitive, high-concurrency, and cost-sensitive scenarios, emphasizing fast response and flexible inference deployment.","created":1772131107,"cutoff":null,"hf":null,"ctx":262144,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":0.1,"out":0.4},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","minimal"]},"moderated":false,"providers":[{"name":"Seed","quant":"fp8","in":0.1,"out":0.4,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance/seedance-1-5-pro","name":"ByteDance: Seedance 1.5 Pro","org":"ByteDance","slug":"seedance-1-5-pro","matchKey":"seedance15pro","blurb":"ByteDance's next-generation audio-visual generation model with a 4.5B parameter Dual-Branch Diffusion Transformer architecture.","created":1774277608,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance/seedance-2.0","name":"ByteDance: Seedance 2.0","org":"ByteDance","slug":"seedance-2.0","matchKey":"seedance20","blurb":"Seedance 2.0 is a video generation model from ByteDance.","created":1776211362,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video","audio"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance/seedance-2.0-fast","name":"ByteDance: Seedance 2.0 Fast","org":"ByteDance","slug":"seedance-2.0-fast","matchKey":"seedance20fast","blurb":"Seedance 2.0 Fast is a video generation model from ByteDance.","created":1776211362,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video","audio"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance/seedance-2.0-mini","name":"ByteDance: Seedance 2.0 Mini","org":"ByteDance","slug":"seedance-2.0-mini","matchKey":"seedance20mini","blurb":"Seedance 2.0 Mini is a video generation model from ByteDance.","created":1786552600,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video","audio"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance/seedance-2.5","name":"ByteDance: Seedance 2.5","org":"ByteDance","slug":"seedance-2.5","matchKey":"seedance25","blurb":"Seedance 2.5 is a video generation model from ByteDance.","created":1786141253,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video","audio"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Seed","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"bytedance/ui-tars-1.5-7b","name":"ByteDance: UI-TARS 7B ","org":"ByteDance","slug":"ui-tars-1.5-7b","matchKey":"uitars157b","blurb":"UI-TARS-1.5 is a multimodal vision-language agent optimized for GUI-based environments, including desktop interfaces, web browsers, mobile systems, and games.","created":1753205056,"cutoff":"2025-01-31","hf":"ByteDance-Seed/UI-TARS-1.5-7B","ctx":128000,"maxOut":2048,"in":["image","text"],"out":["text"],"price":{"in":0.1,"out":0.2},"caps":{"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"bf16","in":0.1,"out":0.2,"cacheRead":0.1,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.1,"n":1},"open":true,"rankable":true,"scored":0},{"id":"cognitivecomputations/dolphin-mistral-24b-venice-edition","name":"Venice: Uncensored","org":"Venice","slug":"dolphin-mistral-24b-venice-edition","matchKey":"dolphinmistral24bveniceedition","blurb":"Venice Uncensored Dolphin Mistral 24B Venice Edition is a fine-tuned variant of Mistral-Small-24B-Instruct-2501, developed by dphn.ai in collaboration with Venice.ai.","created":1752094966,"cutoff":"2024-04-30","hf":"cognitivecomputations/Dolphin-Mistral-24B-Venice-Edition","ctx":128000,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":0.2,"out":0.9},"caps":{"json":true},"reasoning":null,"moderated":false,"providers":[{"name":"Venice","quant":"fp16","in":0.2,"out":0.9,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"cohere/command-r-08-2024","name":"Cohere: Command R (08-2024)","org":"Cohere","slug":"command-r-08-2024","matchKey":"commandr082024","blurb":"command-r-08-2024 is an update of the [Command R](/models/cohere/command-r) with improved performance for multilingual retrieval-augmented generation (RAG) and tool use.","created":1724976000,"cutoff":"2024-03-31","hf":null,"ctx":128000,"maxOut":4000,"in":["text"],"out":["text"],"price":{"in":0.15,"out":0.6},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Cohere","in":0.15,"out":0.6,"uptime":99.4}],"scores":{},"perf":{"uptime":99.4,"n":1},"open":false,"rankable":true,"scored":0},{"id":"cohere/command-r-plus-08-2024","name":"Cohere: Command R+ (08-2024)","org":"Cohere","slug":"command-r-plus-08-2024","matchKey":"commandrplus082024","blurb":"command-r-plus-08-2024 is an update of the [Command R+](/models/cohere/command-r-plus) with roughly 50% higher throughput and 25% lower latencies as compared to the previous Command R+ version, while keeping the hardware","created":1724976000,"cutoff":"2024-03-31","hf":null,"ctx":128000,"maxOut":4000,"in":["text"],"out":["text"],"price":{"in":2.5,"out":10},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Cohere","in":2.5,"out":10,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"cohere/command-r7b-12-2024","name":"Cohere: Command R7B (12-2024)","org":"Cohere","slug":"command-r7b-12-2024","matchKey":"commandr7b122024","blurb":"Command R7B (12-2024) is a small, fast update of the Command R+ model, delivered in December 2024.","created":1734158152,"cutoff":"2024-08-31","hf":null,"ctx":128000,"maxOut":4000,"in":["text"],"out":["text"],"price":{"in":0.0375,"out":0.15},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"Cohere","in":0.0375,"out":0.15,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"deepgram/nova-3","name":"Deepgram: Nova-3","org":"Deepgram","slug":"nova-3","matchKey":"nova3","blurb":"Deepgram Nova-3 general-purpose speech-to-text model with monolingual and multilingual transcription support.","created":1784075761,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":71.666667,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Deepgram","in":71.666667,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"deepseek/deepseek-r1-distill-llama-70b","name":"DeepSeek: R1 Distill Llama 70B","org":"DeepSeek","slug":"deepseek-r1-distill-llama-70b","matchKey":"deepseekr1distillllama70b","blurb":"DeepSeek R1 Distill Llama 70B is a distilled large language model based on [Llama-3.3-70B-Instruct](/meta-llama/llama-3.3-70b-instruct), using outputs from [DeepSeek R1](/deepseek/deepseek-r1).","created":1737663169,"cutoff":"2024-07-31","hf":"deepseek-ai/DeepSeek-R1-Distill-Llama-70B","ctx":8192,"maxOut":7372,"in":["text"],"out":["text"],"price":{"in":0.8,"out":0.8},"caps":{"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Novita","quant":"bf16","in":0.8,"out":0.8,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"fish-audio/transcribe-1","name":"Fish Audio: Transcribe 1","org":"Fish Audio","slug":"transcribe-1","matchKey":"transcribe1","blurb":"Transcribe 1 is a speech-to-text model from Fish Audio.","created":1785353735,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":100,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Fish Audio","in":100,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/chirp-3","name":"Google: Chirp 3","org":"Google","slug":"chirp-3","matchKey":"chirp3","blurb":"Chirp 3 is Google's latest multilingual speech-to-text model.","created":1777997783,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":266.666667,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Google","in":266.666667,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/gemini-3.1-flash-image-preview","name":"Google: Nano Banana 2 (Gemini 3.1 Flash Image Preview)","org":"Google","slug":"gemini-3.1-flash-image-preview","matchKey":"gemini31flashimagepreview","blurb":"Gemini 3.1 Flash Image Preview, a.k.a.","created":1772119558,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["image","text"],"out":["image","text"],"price":{"in":0.5,"out":3,"imageOut":0.00006,"webSearch":0.014},"caps":{"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","minimal"]},"moderated":false,"providers":[{"name":"Google AI Studio","in":0.5,"out":3,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/gemini-3.1-pro-preview-customtools","name":"Google: Gemini 3.1 Pro Preview Custom Tools","org":"Google","slug":"gemini-3.1-pro-preview-customtools","matchKey":"gemini31propreviewcustomtools","blurb":"Gemini 3.1 Pro Preview Custom Tools is a variant of Gemini 3.1 Pro that improves tool selection behavior by preventing overuse of a general bash tool when more efficient third-party...","created":1772045923,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","audio","image","video","file"],"out":["text"],"price":{"in":2,"out":12,"imageIn":0.000002,"audioIn":2,"webSearch":0.014},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Google AI Studio","in":2,"out":12,"cacheRead":0.2,"cacheWrite":0.375,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.2,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/gemma-2-27b-it","name":"Google: Gemma 2 27B","org":"Google","slug":"gemma-2-27b-it","matchKey":"gemma227bit","blurb":"Gemma 2 27B by Google is an open model built from the same research and technology used to create the [Gemini models](/models?q=gemini).","created":1720828800,"cutoff":"2024-06-30","hf":"google/gemma-2-27b-it","ctx":8192,"maxOut":2048,"in":["text"],"out":["text"],"price":{"in":0.65,"out":0.65},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"NextBit","quant":"int4","in":0.65,"out":0.65,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"google/lyria-3-clip-preview","name":"Google: Lyria 3 Clip Preview","org":"Google","slug":"lyria-3-clip-preview","matchKey":"lyria3clippreview","blurb":"30 second duration clips are priced at $0.04 per clip.","created":1774907255,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image"],"out":["text","audio"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google AI Studio","in":0,"out":0,"uptime":99.9}],"scores":{},"perf":{"uptime":99.9,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/lyria-3-pro-preview","name":"Google: Lyria 3 Pro Preview","org":"Google","slug":"lyria-3-pro-preview","matchKey":"lyria3propreview","blurb":"Full-length songs are priced at $0.08 per song.","created":1774907286,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":65536,"in":["text","image"],"out":["text","audio"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google AI Studio","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/veo-3.1","name":"Google: Veo 3.1","org":"Google","slug":"veo-3.1","matchKey":"veo31","blurb":"Google's state-of-the-art video generation model, built for maximum visual fidelity in final production cuts.","created":1774277148,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/veo-3.1-fast","name":"Google: Veo 3.1 Fast","org":"Google","slug":"veo-3.1-fast","matchKey":"veo31fast","blurb":"Google's mid-tier video generation model balancing speed and quality.","created":1776994666,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"google/veo-3.1-lite","name":"Google: Veo 3.1 Lite","org":"Google","slug":"veo-3.1-lite","matchKey":"veo31lite","blurb":"Google's most cost-effective video generation model, designed for high-volume applications and rapid iteration.","created":1776978818,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Google","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"heygen/avatar-iv","name":"HeyGen: Avatar IV","org":"HeyGen","slug":"avatar-iv","matchKey":"avatariv","blurb":"HeyGen: Avatar IV is an image-to-video model that animates a single photo into an expressive, lip-synced talking-head video.","created":1787595728,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","audio"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"HeyGen","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"ibm-granite/granite-4.0-h-micro","name":"IBM: Granite 4.0 Micro","org":"IBM","slug":"granite-4.0-h-micro","matchKey":"granite40hmicro","blurb":"Granite-4.0-H-Micro is a 3B parameter from the Granite 4 family of models.","created":1760927695,"cutoff":null,"hf":"ibm-granite/granite-4.0-h-micro","ctx":131000,"maxOut":117900,"in":["text"],"out":["text"],"price":{"in":0.017,"out":0.112},"caps":{"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Cloudflare","in":0.017,"out":0.112,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"inception/mercury-2.5","name":"Inception: Mercury 2.5","org":"Inception","slug":"mercury-2.5","matchKey":"mercury25","blurb":"Mercury 2.5 is the fastest reasoning LLM, and the latest diffusion LLM (dLLM) from Inception.","created":1788892137,"cutoff":null,"hf":null,"ctx":260000,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.04,"out":0.15},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","medium","low","none"]},"moderated":false,"providers":[{"name":"Inception","in":0.04,"out":0.15,"cacheRead":0.004,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.004,"n":1},"open":false,"rankable":true,"scored":0},{"id":"inclusionai/ling-3.0-flash-fin","name":"inclusionAI: Ling 3.0 Flash Fin","org":"inclusionAI","slug":"ling-3.0-flash-fin","matchKey":"ling30flashfin","blurb":"Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total.","created":1787846290,"cutoff":null,"hf":null,"ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.06,"out":0.18,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp4","in":0.06,"out":0.18,"cacheRead":0.012,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.012,"n":1},"open":false,"rankable":true,"scored":0},{"id":"inclusionai/ling-3.0-flash-vl","name":"inclusionAI: Ling 3.0 Flash VL","org":"inclusionAI","slug":"ling-3.0-flash-vl","matchKey":"ling30flashvl","blurb":"Ling 3.0 Flash VL builds on Ling 3.0 Flash (124B total / 5.5B active MoE from InclusionAI), further strengthening its language capabilities while adding native visual perception and advanced visual...","created":1789056114,"cutoff":null,"hf":"inclusionAI/Ling-3.0-flash-VL","ctx":131072,"maxOut":32768,"in":["text","image","video"],"out":["text"],"price":{"in":0.06,"out":0.18,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp16","in":0.06,"out":0.18,"cacheRead":0.012,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.012,"n":1},"open":true,"rankable":true,"scored":0},{"id":"inference-net/schematron-v2-small","name":"Inference.net: Schematron V2 Small","org":"Inference.net","slug":"schematron-v2-small","matchKey":"schematronv2small","blurb":"Schematron V2 Small is a 3B-parameter HTML-to-JSON extraction model from Inference.net.","created":1789176933,"cutoff":null,"hf":"inference-net/schematron-v2-llama-3.2-3b","ctx":128000,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":0.05,"out":0.23},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"InferenceNet","in":0.05,"out":0.23,"cacheRead":0.05,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.05,"n":1},"open":true,"rankable":true,"scored":0},{"id":"inference-net/schematron-v2-turbo","name":"Inference.net: Schematron V2 Turbo","org":"Inference.net","slug":"schematron-v2-turbo","matchKey":"schematronv2turbo","blurb":"Schematron V2 Turbo is a 3B-parameter HTML-to-JSON extraction model from Inference.net.","created":1789176949,"cutoff":null,"hf":"inference-net/schematron-v2-granite-4.0-h-micro","ctx":128000,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":0.03,"out":0.15},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"InferenceNet","in":0.03,"out":0.15,"cacheRead":0.03,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.03,"n":1},"open":true,"rankable":true,"scored":0},{"id":"krea/krea-2-medium-turbo","name":"Krea: Krea 2 Medium Turbo","org":"Krea","slug":"krea-2-medium-turbo","matchKey":"krea2mediumturbo","blurb":"Krea 2 Medium Turbo is a distilled, speed-focused variant of Krea 2 Medium from Krea.","created":1784574923,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000359281437125749},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Krea","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"kwaipilot/kat-coder-pro-v2","name":"Kwaipilot: KAT-Coder-Pro V2","org":"Kwaipilot","slug":"kat-coder-pro-v2","matchKey":"katcoderprov2","blurb":"KAT-Coder-Pro V2 is the latest high-performance model in KwaiKAT’s KAT-Coder series, designed for complex enterprise-grade software engineering and SaaS integration.","created":1774649310,"cutoff":null,"hf":null,"ctx":262144,"maxOut":144000,"in":["text"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","quant":"fp8","in":0.3,"out":1.2,"cacheRead":0.06}],"scores":{},"perf":{"cache":0.06,"n":1},"open":false,"rankable":true,"scored":0},{"id":"kwaipilot/kat-coder-pro-v2.5","name":"Kwaipilot: KAT-Coder-Pro V2.5","org":"Kwaipilot","slug":"kat-coder-pro-v2.5","matchKey":"katcoderprov25","blurb":"KAT-Coder-Pro V2.5 is a flagship-level Agentic Coding model that can directly hand over an entire issue or an entire business workflow to it, allowing it to autonomously locate and make...","created":1783714589,"cutoff":null,"hf":null,"ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.74,"out":2.96},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","quant":"fp8","in":0.74,"out":2.96,"cacheRead":0.15}],"scores":{},"perf":{"cache":0.15,"n":1},"open":false,"rankable":true,"scored":0},{"id":"kwaivgi/kling-v3.0-pro","name":"Kling: Video v3.0 Pro","org":"Kling","slug":"kling-v3.0-pro","matchKey":"klingv30pro","blurb":"Kling v3.0 Pro is Kuaishou's premium video generation model, offering higher visual quality than the Standard tier.","created":1777496206,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"kwaivgi/kling-v3.0-std","name":"Kling: Video v3.0 Standard","org":"Kling","slug":"kling-v3.0-std","matchKey":"klingv30std","blurb":"Kling v3.0 Standard is a video generation model from Kuaishou.","created":1777496205,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"kwaivgi/kling-video-o1","name":"Kling: Video O1","org":"Kling","slug":"kling-video-o1","matchKey":"klingvideoo1","blurb":"Kling Video O1 is a video generation model from Kuaishou.","created":1776704777,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"AtlasCloud","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"mancer/weaver","name":"Mancer: Weaver (alpha)","org":"Mancer","slug":"weaver","matchKey":"weaver","blurb":"An attempt to recreate Claude-style verbosity, but don't expect the same level of coherence or memory.","created":1690934400,"cutoff":"2023-06-30","hf":null,"ctx":8000,"maxOut":6000,"in":["text"],"out":["text"],"price":{"in":0.4,"out":0.75},"caps":{"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Mancer 2","quant":"fp8","in":0.4,"out":0.75,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"meituan/longcat-2.0","name":"Meituan: LongCat 2.0","org":"Meituan","slug":"longcat-2.0","matchKey":"longcat20","blurb":"LongCat 2.0 is a sparse mixture-of-experts language model from Meituan, with 48B active parameters out of 1.6T total.","created":1784554658,"cutoff":null,"hf":"meituan-longcat/LongCat-2.0","ctx":1048756,"maxOut":262144,"in":["text"],"out":["text"],"price":{"in":0.3,"out":1.2},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"AtlasCloud","quant":"fp8","in":0.3,"out":1.2,"cacheRead":0.006,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.006,"n":1},"open":true,"rankable":true,"scored":0},{"id":"meta-llama/llama-guard-4-12b","name":"Meta: Llama Guard 4 12B","org":"Meta","slug":"llama-guard-4-12b","matchKey":"llamaguard412b","blurb":"Llama Guard 4 is a Llama 4 Scout-derived multimodal pretrained model, fine-tuned for content safety classification.","created":1745975193,"cutoff":"2024-08-31","hf":"meta-llama/Llama-Guard-4-12B","ctx":163840,"maxOut":16384,"in":["image","text"],"out":["text"],"price":{"in":0.18,"out":0.18},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":0.18,"out":0.18,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"meta/muse-spark-1.2-contributor","name":"Meta: Muse Spark 1.2 Contributor","org":"Meta","slug":"muse-spark-1.2-contributor","matchKey":"musespark12contributor","blurb":"Muse Spark 1.2 contributor tier is a reasoning model from Meta designed for developers who want to start building at an even lower cost.","created":1787336476,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":943718,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.1,"out":0.2,"webSearch":0.0025},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"Meta","in":0.1,"out":0.2,"cacheRead":0.002,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.002,"n":1},"open":false,"rankable":true,"scored":0},{"id":"meta/muse-spark-1.3-contributor","name":"Meta: Muse Spark 1.3 Contributor","org":"Meta","slug":"muse-spark-1.3-contributor","matchKey":"musespark13contributor","blurb":"Muse Spark 1.3 Contributor is the cost-efficient contributor tier of Meta’s multimodal reasoning model for experimentation, learning, and early-stage agentic, multi-agent, and coding workflows.","created":1788381519,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":943718,"in":["text","image","video","file","audio"],"out":["text"],"price":{"in":0.1,"out":0.2,"webSearch":0.0025},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high","medium","low","minimal"]},"moderated":true,"providers":[{"name":"Meta","in":0.1,"out":0.2,"cacheRead":0.002,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.002,"n":1},"open":false,"rankable":true,"scored":0},{"id":"meta/muse-voice-transcribe-1.0","name":"Meta: Muse Voice Transcribe 1.0","org":"Meta","slug":"muse-voice-transcribe-1.0","matchKey":"musevoicetranscribe10","blurb":"Muse Voice Transcribe 1.0 is a synchronous speech-to-text model from Meta.","created":1789146931,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":50,"out":0},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"Meta","in":50,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"microsoft/mai-image-2.5-pro","name":"Microsoft AI: MAI-Image-2.5 Pro","org":"Microsoft AI","slug":"mai-image-2.5-pro","matchKey":"maiimage25pro","blurb":"Microsoft AI's MAI-Image-2.5 is a high-quality image generation model available via Azure AI Foundry.","created":1784827701,"cutoff":null,"hf":null,"ctx":4096,"maxOut":1024,"in":["text","image"],"out":["image"],"price":{"in":5,"out":0,"imageOut":0.000108},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":5,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"microsoft/mai-image-2.6-flash","name":"Microsoft AI: MAI-Image-2.6 Flash","org":"Microsoft AI","slug":"mai-image-2.6-flash","matchKey":"maiimage26flash","blurb":"MAI-Image-2.6 Flash is the lower-latency, lower-cost member of the [MAI-Image-2.6](/microsoft/mai-image-2.6) family from Microsoft AI, built for latency-sensitive, high-throughput production image generation and editing ","created":1788550735,"cutoff":null,"hf":null,"ctx":4096,"maxOut":1024,"in":["text","image"],"out":["image"],"price":{"in":1.75,"out":0,"imageOut":0.000019},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":1.75,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"microsoft/mai-transcribe-1.5","name":"Microsoft AI: MAI-Transcribe 1.5","org":"Microsoft AI","slug":"mai-transcribe-1.5","matchKey":"maitranscribe15","blurb":"MAI-Transcribe 1.5 is a multilingual speech-to-text model from Microsoft AI.","created":1780425095,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":360000,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":360000,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"microsoft/mai-transcribe-2","name":"Microsoft AI: MAI-Transcribe 2","org":"Microsoft AI","slug":"mai-transcribe-2","matchKey":"maitranscribe2","blurb":"MAI-Transcribe 2 is a multilingual speech-to-text model from Microsoft AI, ranked #1 on the FLEURS multilingual benchmark.","created":1788396586,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":100000,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Azure","in":100000,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"microsoft/wizardlm-2-8x22b","name":"WizardLM-2 8x22B","org":"microsoft","slug":"wizardlm-2-8x22b","matchKey":"wizardlm28x22b","blurb":"WizardLM-2 8x22B is Microsoft AI's most advanced Wizard model.","created":1713225600,"cutoff":"2024-04-30","hf":"microsoft/WizardLM-2-8x22B","ctx":65535,"maxOut":8000,"in":["text"],"out":["text"],"price":{"in":0.62,"out":0.62},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Novita","quant":"bf16","in":0.62,"out":0.62,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"minimax/hailuo-3","name":"MiniMax: H3","org":"MiniMax","slug":"hailuo-3","matchKey":"hailuo3","blurb":"MiniMax H3 is a lightweight, open-weights video generation model from MiniMax.","created":1785366648,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video","audio"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Minimax","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"minimax/hailuo-3-max","name":"MiniMax: H3 Max","org":"MiniMax","slug":"hailuo-3-max","matchKey":"hailuo3max","blurb":"MiniMax H3 Max is a video-generation model from MiniMax, jointly released with fal.ai.","created":1788313310,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Minimax","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"minimax/minimax-01","name":"MiniMax: MiniMax-01","org":"MiniMax","slug":"minimax-01","matchKey":"minimax01","blurb":"MiniMax-01 is a combines MiniMax-Text-01 for text generation and MiniMax-VL-01 for image understanding.","created":1736915462,"cutoff":"2024-03-31","hf":"MiniMaxAI/MiniMax-Text-01","ctx":1000192,"maxOut":900172,"in":["text","image"],"out":["text"],"price":{"in":0.2,"out":1.1},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Minimax","in":0.2,"out":1.1,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"mistralai/voxtral-mini-3b-2507","name":"Mistral: Voxtral Mini 3B 2507","org":"Mistral","slug":"voxtral-mini-3b-2507","matchKey":"voxtralmini3b2507","blurb":"Voxtral Mini 3B 2507 is a speech and audio understanding model from Mistral AI.","created":1786653980,"cutoff":null,"hf":"mistralai/Voxtral-Mini-3B-2507","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":16.6667,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":16.6667,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"mistralai/voxtral-small-24b-2507-stt","name":"Mistral: Voxtral Small 24B 2507 STT","org":"Mistral","slug":"voxtral-small-24b-2507-stt","matchKey":"voxtralsmall24b2507stt","blurb":"Voxtral Small 24B 2507 STT is a speech transcription model from Mistral AI.","created":1786654002,"cutoff":null,"hf":"mistralai/Voxtral-Small-24B-2507","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":50,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":50,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"morph/morph-v3-fast","name":"Morph: Morph V3 Fast","org":"Morph","slug":"morph-v3-fast","matchKey":"morphv3fast","blurb":"Morph's fastest apply model for code edits.","created":1751910002,"cutoff":null,"hf":null,"ctx":81920,"maxOut":38000,"in":["text"],"out":["text"],"price":{"in":0.8,"out":1.2},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Morph","in":0.8,"out":1.2,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"morph/morph-v3-large","name":"Morph: Morph V3 Large","org":"Morph","slug":"morph-v3-large","matchKey":"morphv3large","blurb":"Morph's high-accuracy apply model for complex code edits.","created":1751910858,"cutoff":null,"hf":null,"ctx":262144,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.9,"out":1.9},"caps":{"json":true,"schema":true,"logprobs":true},"reasoning":null,"moderated":false,"providers":[{"name":"Morph","in":0.9,"out":1.9,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"nousresearch/hermes-3-llama-3.1-405b","name":"Nous: Hermes 3 405B Instruct","org":"Nous","slug":"hermes-3-llama-3.1-405b","matchKey":"hermes3llama31405b","blurb":"Hermes 3 is a generalist language model with many improvements over Hermes 2, including advanced agentic capabilities, much better roleplaying, reasoning, multi-turn conversation, long context coherence, and improvements","created":1723766400,"cutoff":"2023-12-31","hf":"NousResearch/Hermes-3-Llama-3.1-405B","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":1,"out":1},"caps":{"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":1,"out":1,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"nousresearch/hermes-3-llama-3.1-70b","name":"Nous: Hermes 3 70B Instruct","org":"Nous","slug":"hermes-3-llama-3.1-70b","matchKey":"hermes3llama3170b","blurb":"Hermes 3 is a generalist language model with many improvements over [Hermes 2](/models/nousresearch/nous-hermes-2-mistral-7b-dpo), including advanced agentic capabilities, much better roleplaying, reasoning, multi-turn c","created":1723939200,"cutoff":"2023-12-31","hf":"NousResearch/Hermes-3-Llama-3.1-70B","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.7,"out":0.7},"caps":{"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","quant":"fp8","in":0.7,"out":0.7,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"nvidia/nemotron-3.5-asr-streaming-multilingual-0.6b","name":"NVIDIA: Nemotron 3.5 ASR Streaming Multilingual 0.6B","org":"NVIDIA","slug":"nemotron-3.5-asr-streaming-multilingual-0.6b","matchKey":"nemotron35asrstreamingmultilingual06b","blurb":"Nemotron 3.5 ASR Streaming Multilingual 0.6B is a speech recognition model from NVIDIA.","created":1786654371,"cutoff":null,"hf":"nvidia/Nemotron-3.5-ASR-Streaming-Multilingual-0.6b","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":3.33,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","in":3.33,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"nvidia/nemotron-3.5-content-safety","name":"NVIDIA: Nemotron 3.5 Content Safety","org":"NVIDIA","slug":"nemotron-3.5-content-safety","matchKey":"nemotron35contentsafety","blurb":"NVIDIA Nemotron 3.5 Content Safety is a compact 4B-parameter multimodal guardrail model from NVIDIA, fine-tuned from Google Gemma-3-4B.","created":1780581864,"cutoff":null,"hf":"nvidia/Nemotron-3.5-Content-Safety","ctx":131072,"maxOut":117964,"in":["text","image"],"out":["text"],"price":{"in":0.2,"out":0.2,"alt":{"free":{"in":0,"out":0}}},"caps":{"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"DeepInfra","quant":"bf16","in":0.2,"out":0.2,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"nvidia/parakeet-tdt-0.6b-v3","name":"NVIDIA: Parakeet TDT 0.6B v3","org":"NVIDIA","slug":"parakeet-tdt-0.6b-v3","matchKey":"parakeettdt06bv3","blurb":"Parakeet TDT 0.6B v3 is NVIDIA's 600M-parameter multilingual speech-to-text model built on the FastConformer-TDT architecture.","created":1779848335,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":25,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Together","in":25,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-3.5-turbo-instruct","name":"OpenAI: GPT-3.5 Turbo Instruct","org":"OpenAI","slug":"gpt-3.5-turbo-instruct","matchKey":"gpt35turboinstruct","blurb":"This model is a variant of GPT-3.5 Turbo tuned for instructional prompts and omitting chat-related optimizations.","created":1695859200,"cutoff":"2021-09-30","hf":null,"ctx":4095,"maxOut":3685,"in":["text"],"out":["text"],"price":{"in":1.5,"out":2},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":1.5,"out":2,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-4o-2024-11-20","name":"OpenAI: GPT-4o (2024-11-20)","org":"OpenAI","slug":"gpt-4o-2024-11-20","matchKey":"gpt4o20241120","blurb":"The 2024-11-20 version of GPT-4o offers a leveled-up creative writing ability with more natural, engaging, and tailored writing to improve relevance & readability.","created":1732127594,"cutoff":"2023-10-31","hf":null,"ctx":128000,"maxOut":16384,"in":["text","image","file"],"out":["text"],"price":{"in":2.5,"out":10},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":2.5,"out":10,"cacheRead":1.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":1.25,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-4o-mini-transcribe","name":"OpenAI: GPT-4o Mini Transcribe","org":"OpenAI","slug":"gpt-4o-mini-transcribe","matchKey":"gpt4ominitranscribe","blurb":"GPT-4o Mini Transcribe is OpenAI's smaller, cost-efficient speech-to-text model built on GPT-4o Mini audio capabilities.","created":1777658151,"cutoff":null,"hf":null,"ctx":128000,"maxOut":115200,"in":["audio"],"out":["transcription"],"price":{"in":1.25,"out":5},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":1.25,"out":5,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-4o-transcribe","name":"OpenAI: GPT-4o Transcribe","org":"OpenAI","slug":"gpt-4o-transcribe","matchKey":"gpt4otranscribe","blurb":"GPT-4o Transcribe is OpenAI's high-quality speech-to-text model built on GPT-4o audio capabilities.","created":1777332895,"cutoff":null,"hf":null,"ctx":128000,"maxOut":115200,"in":["audio"],"out":["transcription"],"price":{"in":2.5,"out":10},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":2.5,"out":10,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-5-image","name":"OpenAI: GPT-5 Image","org":"OpenAI","slug":"gpt-5-image","matchKey":"gpt5image","blurb":"[GPT-5](https://openrouter.ai/openai/gpt-5) Image combines OpenAI's GPT-5 model with state-of-the-art image generation capabilities.","created":1760447986,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["image","text","file"],"out":["image","text"],"price":{"in":10,"out":10,"imageOut":0.00004,"webSearch":0.01},"caps":{"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":10,"out":10,"cacheRead":1.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":1.25,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-5-image-mini","name":"OpenAI: GPT-5 Image Mini","org":"OpenAI","slug":"gpt-5-image-mini","matchKey":"gpt5imagemini","blurb":"GPT-5 Image Mini combines OpenAI's advanced language capabilities, powered by [GPT-5 Mini](https://openrouter.ai/openai/gpt-5-mini), with GPT Image 1 Mini for efficient image generation.","created":1760624583,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["file","image","text"],"out":["image","text"],"price":{"in":2.5,"out":2,"imageOut":0.000008,"webSearch":0.01},"caps":{"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":2.5,"out":2,"cacheRead":0.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.25,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-5.1-codex-max","name":"OpenAI: GPT-5.1-Codex-Max","org":"OpenAI","slug":"gpt-5.1-codex-max","matchKey":"gpt51codexmax","blurb":"GPT-5.1-Codex-Max is OpenAI’s latest agentic coding model, designed for long-running, high-context software development tasks.","created":1764878934,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["text","image"],"out":["text"],"price":{"in":1.25,"out":10,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"Azure","in":1.25,"out":10,"cacheRead":0.125,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.125,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-5.4-image-2","name":"OpenAI: GPT-5.4 Image 2","org":"OpenAI","slug":"gpt-5.4-image-2","matchKey":"gpt54image2","blurb":"[GPT-5.4](https://openrouter.ai/openai/gpt-5.4) Image 2 combines OpenAI's GPT-5.4 model with state-of-the-art image generation capabilities from GPT Image 2.","created":1776797528,"cutoff":null,"hf":null,"ctx":272000,"maxOut":128000,"in":["image","text","file"],"out":["image","text"],"price":{"in":8,"out":15,"imageOut":0.00003,"webSearch":0.01},"caps":{"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","high","medium","low","none"]},"moderated":true,"providers":[{"name":"OpenAI","in":8,"out":15,"cacheRead":2,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":2,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-audio","name":"OpenAI: GPT Audio","org":"OpenAI","slug":"gpt-audio","matchKey":"gptaudio","blurb":"The gpt-audio model is OpenAI's first generally available audio model.","created":1768862569,"cutoff":null,"hf":null,"ctx":128000,"maxOut":16384,"in":["text","audio"],"out":["text","audio"],"price":{"in":2.5,"out":10,"audioIn":32,"audioOut":64},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":2.5,"out":10,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-audio-mini","name":"OpenAI: GPT Audio Mini","org":"OpenAI","slug":"gpt-audio-mini","matchKey":"gptaudiomini","blurb":"A cost-efficient version of GPT Audio.","created":1768859419,"cutoff":null,"hf":null,"ctx":128000,"maxOut":16384,"in":["text","audio"],"out":["text","audio"],"price":{"in":0.6,"out":2.4,"audioIn":0.6,"audioOut":2.4},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":0.6,"out":2.4,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-chat-latest","name":"OpenAI: GPT Chat Latest","org":"OpenAI","slug":"gpt-chat-latest","matchKey":"gptchatlatest","blurb":"GPT Chat Latest points to OpenAI's stable API alias `chat-latest` that always resolves to the latest Instant chat model used in ChatGPT.","created":1778000212,"cutoff":null,"hf":null,"ctx":400000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":30,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":5,"out":30,"cacheRead":0.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.5,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/gpt-oss-safeguard-20b","name":"OpenAI: gpt-oss-safeguard-20b","org":"OpenAI","slug":"gpt-oss-safeguard-20b","matchKey":"gptosssafeguard20b","blurb":"gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b.","created":1761752836,"cutoff":null,"hf":"openai/gpt-oss-safeguard-20b","ctx":131072,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.075,"out":0.3},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Groq","in":0.075,"out":0.3,"cacheRead":0.0375,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.0375,"n":1},"open":true,"rankable":true,"scored":0},{"id":"openai/gpt-transcribe","name":"OpenAI: GPT Transcribe","org":"OpenAI","slug":"gpt-transcribe","matchKey":"gpttranscribe","blurb":"GPT Transcribe is a high-accuracy speech-to-text model from OpenAI.","created":1785973897,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":75,"out":0},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":75,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/o1-pro","name":"OpenAI: o1-pro","org":"OpenAI","slug":"o1-pro","matchKey":"o1pro","blurb":"The o1 series of models are trained with reinforcement learning to think before they answer and perform complex reasoning.","created":1742423211,"cutoff":"2023-10-31","hf":null,"ctx":200000,"maxOut":100000,"in":["text","image","file"],"out":["text"],"price":{"in":150,"out":600,"webSearch":0.01},"caps":{"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":150,"out":600,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/o3-pro","name":"OpenAI: o3 Pro","org":"OpenAI","slug":"o3-pro","matchKey":"o3pro","blurb":"The o-series of models are trained with reinforcement learning to think before they answer and perform complex reasoning.","created":1749598352,"cutoff":"2024-06-30","hf":null,"ctx":200000,"maxOut":100000,"in":["text","file","image"],"out":["text"],"price":{"in":20,"out":80,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[{"name":"OpenAI","in":20,"out":80,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/o4-mini-high","name":"OpenAI: o4 Mini High","org":"OpenAI","slug":"o4-mini-high","matchKey":"o4minihigh","blurb":"OpenAI o4-mini-high is the same model as [o4-mini](/openai/o4-mini) with reasoning_effort set to high.","created":1744824212,"cutoff":"2024-06-30","hf":null,"ctx":200000,"maxOut":100000,"in":["image","text","file"],"out":["text"],"price":{"in":1.1,"out":4.4,"webSearch":0.01},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":["high"]},"moderated":true,"providers":[{"name":"OpenAI","in":1.1,"out":4.4,"cacheRead":0.275,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.275,"n":1},"open":false,"rankable":true,"scored":0},{"id":"openai/whisper-1","name":"OpenAI: Whisper 1","org":"OpenAI","slug":"whisper-1","matchKey":"whisper1","blurb":"Whisper is OpenAI's open-source automatic speech recognition model, available via API as `whisper-1`.","created":1777332905,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":100,"out":0},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"OpenAI","in":100,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"perceptron/perceptron-mk1","name":"Perceptron: Perceptron Mk1","org":"Perceptron","slug":"perceptron-mk1","matchKey":"perceptronmk1","blurb":"Perceptron Mk1 (Mark One) is Perceptron's highest-quality vision-language model for video and embodied reasoning.** It accepts image and video inputs paired with natural language queries, and produces detailed visual und","created":1778597029,"cutoff":null,"hf":null,"ctx":32768,"maxOut":8192,"in":["text","image","video"],"out":["text"],"price":{"in":0.15,"out":1.5},"caps":{"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Perceptron","in":0.15,"out":1.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"perplexity/sonar","name":"Perplexity: Sonar","org":"Perplexity","slug":"sonar","matchKey":"sonar","blurb":"Sonar is lightweight, affordable, fast, and simple to use — now featuring citations and the ability to customize sources.","created":1738013808,"cutoff":null,"hf":null,"ctx":127072,"maxOut":114364,"in":["text","image"],"out":["text"],"price":{"in":1,"out":1,"webSearch":0.005},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Perplexity","in":1,"out":1,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"perplexity/sonar-deep-research","name":"Perplexity: Sonar Deep Research","org":"Perplexity","slug":"sonar-deep-research","matchKey":"sonardeepresearch","blurb":"Sonar Deep Research is a research-focused model designed for multi-step retrieval, synthesis, and reasoning across complex topics.","created":1741311246,"cutoff":null,"hf":null,"ctx":128000,"maxOut":115200,"in":["text"],"out":["text"],"price":{"in":2,"out":8,"webSearch":0.005},"caps":{"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Perplexity","in":2,"out":8,"uptime":99.9}],"scores":{},"perf":{"uptime":99.9,"n":1},"open":false,"rankable":true,"scored":0},{"id":"perplexity/sonar-pro-search","name":"Perplexity: Sonar Pro Search","org":"Perplexity","slug":"sonar-pro-search","matchKey":"sonarprosearch","blurb":"Exclusively available on the OpenRouter API, Sonar Pro's new Pro Search mode is Perplexity's most advanced agentic search system.","created":1761854366,"cutoff":null,"hf":null,"ctx":200000,"maxOut":8000,"in":["text","image"],"out":["text"],"price":{"in":3,"out":15,"webSearch":0.018},"caps":{"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Perplexity","in":3,"out":15,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"poolside/laguna-s-2.1","name":"Poolside: Laguna S 2.1","org":"Poolside","slug":"laguna-s-2.1","matchKey":"lagunas21","blurb":"Laguna S 2.1 is the latest coding agent model from [Poolside](<https://poolside.ai/>).","created":1784652683,"cutoff":null,"hf":"poolside/Laguna-S-2.1","ctx":1048576,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":0.09,"out":0.18,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Poolside","quant":"fp4","in":0.09,"out":0.18,"cacheRead":0.009,"uptime":99.9}],"scores":{},"perf":{"uptime":99.9,"cache":0.009,"n":1},"open":true,"rankable":true,"scored":0},{"id":"poolside/laguna-xs-2.1","name":"Poolside: Laguna XS 2.1","org":"Poolside","slug":"laguna-xs-2.1","matchKey":"lagunaxs21","blurb":"Laguna XS 2.1 is the latest coding agent model in the 33B-A3B category from [Poolside](https://poolside.ai/) and a step forward from their Laguna XS.2 model (released in April 2026).","created":1783002429,"cutoff":null,"hf":"poolside/Laguna-XS-2.1","ctx":262144,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.06,"out":0.12,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Poolside","quant":"fp8","in":0.06,"out":0.12,"cacheRead":0.03,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.03,"n":1},"open":true,"rankable":true,"scored":0},{"id":"prism-ml/ternary-bonsai-2-27b","name":"PrismML: Ternary Bonsai 2 27B","org":"PrismML","slug":"ternary-bonsai-2-27b","matchKey":"ternarybonsai227b","blurb":"Bonsai 2 27B is a 27B-parameter reasoning model from PrismML derived from Qwen3.8-27B.","created":1789754046,"cutoff":null,"hf":"prism-ml/Ternary-Bonsai-2-27B-gguf","ctx":262144,"maxOut":32768,"in":["text","image"],"out":["text"],"price":{"in":0.075,"out":0.5},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["xhigh","medium"]},"moderated":false,"providers":[{"name":"Darkbloom","quant":"int4","in":0.075,"out":0.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen-2.5-7b-instruct","name":"Qwen: Qwen2.5 7B Instruct","org":"Qwen","slug":"qwen-2.5-7b-instruct","matchKey":"qwen257binstruct","blurb":"Qwen2.5 7B is the latest series of Qwen large language models.","created":1729036800,"cutoff":"2024-06-30","hf":"Qwen/Qwen2.5-7B-Instruct","ctx":32768,"maxOut":29491,"in":["text"],"out":["text"],"price":{"in":0.1,"out":0.2},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Phala","in":0.1,"out":0.2,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen-image-3","name":"Qwen: Qwen Image 3","org":"Qwen","slug":"qwen-image-3","matchKey":"qwenimage3","blurb":"Qwen Image 3 is a unified image generation and editing model from Qwen.","created":1785894548,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000718562874251497,"imageIn":0.003},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen-image-3-pro","name":"Qwen: Qwen Image 3 Pro","org":"Qwen","slug":"qwen-image-3-pro","matchKey":"qwenimage3pro","blurb":"Qwen Image 3 Pro is an image generation and editing model from Qwen.","created":1785894548,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000958083832335329,"imageIn":0.003},"caps":{"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen-plus","name":"Qwen: Qwen-Plus","org":"Qwen","slug":"qwen-plus","matchKey":"qwenplus","blurb":"Qwen-Plus, based on the Qwen2.5 foundation model, is a 131K context model with a balanced performance, speed, and cost combination.","created":1738409840,"cutoff":"2025-03-31","hf":null,"ctx":1000000,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.26,"out":0.78},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0.26,"out":0.78,"cacheRead":0.052,"cacheWrite":0.325,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.052,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen-plus-2025-07-28","name":"Qwen: Qwen Plus 0728","org":"Qwen","slug":"qwen-plus-2025-07-28","matchKey":"qwenplus20250728","blurb":"Qwen Plus 0728, based on the Qwen3 foundation model, is a 1 million context hybrid reasoning model with a balanced performance, speed, and cost combination.","created":1757347599,"cutoff":"2025-03-31","hf":null,"ctx":1000000,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.26,"out":0.78},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.26,"out":0.78,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3-30b-a3b-thinking-2507","name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"Qwen","slug":"qwen3-30b-a3b-thinking-2507","matchKey":"qwen330ba3bthinking2507","blurb":"Qwen3-30B-A3B-Thinking-2507 is a 30B parameter Mixture-of-Experts reasoning model optimized for complex tasks requiring extended multi-step thinking.","created":1756399192,"cutoff":"2025-06-30","hf":"Qwen/Qwen3-30B-A3B-Thinking-2507","ctx":81920,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0.2,"out":2.4},"caps":{"tools":true,"json":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.2,"out":2.4,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-asr-0.6b","name":"Qwen: Qwen3 ASR 0.6B","org":"Qwen","slug":"qwen3-asr-0.6b","matchKey":"qwen3asr06b","blurb":"Qwen3 ASR 0.6B is a compact automatic speech recognition model from Qwen.","created":1786591833,"cutoff":null,"hf":"Qwen/Qwen3-ASR-0.6B","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":3.33,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","in":3.33,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-asr-1.7b","name":"Qwen: Qwen3 ASR 1.7B","org":"Qwen","slug":"qwen3-asr-1.7b","matchKey":"qwen3asr17b","blurb":"Qwen3 ASR 1.7B is an automatic speech recognition model from Qwen.","created":1786592646,"cutoff":null,"hf":"Qwen/Qwen3-ASR-1.7B","ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":7.5,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"DeepInfra","in":7.5,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-asr-flash-2026-02-10","name":"Qwen: Qwen3 ASR Flash","org":"Qwen","slug":"qwen3-asr-flash-2026-02-10","matchKey":"qwen3asrflash20260210","blurb":"Qwen3-ASR-Flash is Alibaba's automatic speech recognition service, built on the Qwen3-Omni foundation and trained on tens of millions of hours of multimodal speech data.","created":1778732776,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":35,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":35,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3-coder-flash","name":"Qwen: Qwen3 Coder Flash","org":"Qwen","slug":"qwen3-coder-flash","matchKey":"qwen3coderflash","blurb":"Qwen3 Coder Flash is Alibaba's fast and cost efficient version of their proprietary Qwen3 Coder Plus.","created":1758115536,"cutoff":"2025-06-30","hf":null,"ctx":1000000,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.195,"out":0.975},"caps":{"tools":true,"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0.195,"out":0.975,"cacheRead":0.039,"cacheWrite":0.24375,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.039,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3-coder-plus","name":"Qwen: Qwen3 Coder Plus","org":"Qwen","slug":"qwen3-coder-plus","matchKey":"qwen3coderplus","blurb":"Qwen3 Coder Plus is Alibaba's proprietary version of the Open Source Qwen3 Coder 480B A35B.","created":1758662707,"cutoff":"2025-06-30","hf":null,"ctx":1000000,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.65,"out":3.25},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0.65,"out":3.25,"cacheRead":0.13,"cacheWrite":0.8125,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.13,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3-max-thinking","name":"Qwen: Qwen3 Max Thinking","org":"Qwen","slug":"qwen3-max-thinking","matchKey":"qwen3maxthinking","blurb":"Qwen3-Max-Thinking is the flagship reasoning model in the Qwen3 series, designed for high-stakes cognitive tasks that require deep, multi-step reasoning.","created":1770671901,"cutoff":null,"hf":null,"ctx":262144,"maxOut":65536,"in":["text"],"out":["text"],"price":{"in":0.78,"out":3.9},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.78,"out":3.9,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3-vl-32b-instruct","name":"Qwen: Qwen3 VL 32B Instruct","org":"Qwen","slug":"qwen3-vl-32b-instruct","matchKey":"qwen3vl32binstruct","blurb":"Qwen3-VL-32B-Instruct is a large-scale multimodal vision-language model designed for high-precision understanding and reasoning across text, images, and video.","created":1761231332,"cutoff":null,"hf":"Qwen/Qwen3-VL-32B-Instruct","ctx":131072,"maxOut":32768,"in":["text","image"],"out":["text"],"price":{"in":0.104,"out":0.416},"caps":{"tools":true,"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Alibaba","in":0.104,"out":0.416,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3-vl-8b-thinking","name":"Qwen: Qwen3 VL 8B Thinking","org":"Qwen","slug":"qwen3-vl-8b-thinking","matchKey":"qwen3vl8bthinking","blurb":"Qwen3-VL-8B-Thinking is the reasoning-optimized variant of the Qwen3-VL-8B multimodal model, designed for advanced visual and textual reasoning across complex scenes, documents, and temporal sequences.","created":1760463746,"cutoff":null,"hf":"Qwen/Qwen3-VL-8B-Thinking","ctx":131072,"maxOut":32768,"in":["image","text"],"out":["text"],"price":{"in":0.18,"out":2.1},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.18,"out":2.1,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"qwen/qwen3.5-flash-02-23","name":"Qwen: Qwen3.5-Flash","org":"Qwen","slug":"qwen3.5-flash-02-23","matchKey":"qwen35flash0223","blurb":"The Qwen3.5 native vision-language Flash models are built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency.","created":1772053776,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.065,"out":0.26},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.065,"out":0.26,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3.5-plus-02-15","name":"Qwen: Qwen3.5 Plus 2026-02-15","org":"Qwen","slug":"qwen3.5-plus-02-15","matchKey":"qwen35plus0215","blurb":"The Qwen3.5 native vision-language series Plus models are built on a hybrid architecture that integrates linear attention mechanisms with sparse mixture-of-experts models, achieving higher inference efficiency.","created":1771229416,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":65536,"in":["text","image","video"],"out":["text"],"price":{"in":0.26,"out":1.56},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.26,"out":1.56,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"qwen/qwen3.8-flash","name":"Qwen: Qwen3.8 Flash","org":"Qwen","slug":"qwen3.8-flash","matchKey":"qwen38flash","blurb":"Qwen3.8 Flash is a multimodal reasoning model from Alibaba.","created":1787773060,"cutoff":null,"hf":"Qwen/Qwen3.8-Flash-Next","ctx":1000000,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":0.15,"out":0.47},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Alibaba","in":0.15,"out":0.47,"cacheRead":0.016,"cacheWrite":0.2,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.016,"n":1},"open":true,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-pro","name":"Recraft: Recraft V4 Pro","org":"Recraft","slug":"recraft-v4-pro","matchKey":"recraftv4pro","blurb":"Recraft V4 Pro is an image generation model from Recraft.","created":1778185441,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000598802395209581},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-pro-vector","name":"Recraft: Recraft V4 Pro Vector","org":"Recraft","slug":"recraft-v4-pro-vector","matchKey":"recraftv4provector","blurb":"Recraft V4 Pro Vector is the vector (SVG) variant of Recraft V4 Pro.","created":1778707334,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000718562874251497},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-styles","name":"Recraft: Recraft V4 Styles","org":"Recraft","slug":"recraft-v4-styles","matchKey":"recraftv4styles","blurb":"Recraft V4 Styles is a style-consistent image generation model from Recraft.","created":1787742458,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000838323353293413},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-styles-pro","name":"Recraft: Recraft V4 Styles Pro","org":"Recraft","slug":"recraft-v4-styles-pro","matchKey":"recraftv4stylespro","blurb":"Recraft V4 Styles Pro is a style-consistent image generation model from Recraft.","created":1787742640,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000239520958083832},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-styles-pro-vector","name":"Recraft: Recraft V4 Styles Pro Vector","org":"Recraft","slug":"recraft-v4-styles-pro-vector","matchKey":"recraftv4stylesprovector","blurb":"Recraft V4 Styles Pro Vector is a style-consistent image generation model from Recraft.","created":1787742628,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000287425149700599},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-styles-vector","name":"Recraft: Recraft V4 Styles Vector","org":"Recraft","slug":"recraft-v4-styles-vector","matchKey":"recraftv4stylesvector","blurb":"Recraft V4 Styles Vector is a style-consistent image generation model from Recraft.","created":1787742635,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000119760479041916},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4-vector","name":"Recraft: Recraft V4 Vector","org":"Recraft","slug":"recraft-v4-vector","matchKey":"recraftv4vector","blurb":"Recraft V4 Vector is the vector (SVG) variant of Recraft V4.","created":1778707333,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000191616766467066},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4.1","name":"Recraft: Recraft V4.1","org":"Recraft","slug":"recraft-v4.1","matchKey":"recraftv41","blurb":"Recraft V4.1 is an image generation model from Recraft tuned for high aesthetics.","created":1778707381,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000838323353293413},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4.1-pro-vector","name":"Recraft: Recraft V4.1 Pro Vector","org":"Recraft","slug":"recraft-v4.1-pro-vector","matchKey":"recraftv41provector","blurb":"Recraft V4.1 Pro Vector is the vector (SVG) variant of Recraft V4.1 Pro, tuned for high aesthetics.","created":1778707395,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000718562874251497},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4.1-utility","name":"Recraft: Recraft V4.1 Utility","org":"Recraft","slug":"recraft-v4.1-utility","matchKey":"recraftv41utility","blurb":"Recraft V4.1 Utility is a general-purpose image generation model from Recraft.","created":1778707387,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000838323353293413},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"recraft/recraft-v4.1-vector","name":"Recraft: Recraft V4.1 Vector","org":"Recraft","slug":"recraft-v4.1-vector","matchKey":"recraftv41vector","blurb":"Recraft V4.1 Vector is the vector (SVG) variant of Recraft V4.1, tuned for high aesthetics.","created":1778707392,"cutoff":null,"hf":null,"ctx":65536,"maxOut":58982,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000191616766467066},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Recraft","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"rekaai/reka-edge","name":"Reka Edge","org":"rekaai","slug":"reka-edge","matchKey":"rekaedge","blurb":"Reka Edge is an extremely efficient 7B multimodal vision-language model that accepts image/video+text inputs and generates text outputs.","created":1774026965,"cutoff":null,"hf":"RekaAI/reka-edge-2603","ctx":16384,"maxOut":14745,"in":["image","text","video"],"out":["text"],"price":{"in":0.1,"out":0.1},"caps":{"tools":true,"schema":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Reka","quant":"bf16","in":0.1,"out":0.1,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"relace/relace-apply-3","name":"Relace: Relace Apply 3","org":"Relace","slug":"relace-apply-3","matchKey":"relaceapply3","blurb":"Relace Apply 3 is a specialized code-patching LLM that merges AI-suggested edits straight into your source files.","created":1758891572,"cutoff":null,"hf":null,"ctx":256000,"maxOut":128000,"in":["text"],"out":["text"],"price":{"in":0.85,"out":1.25},"caps":{"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Relace","quant":"fp8","in":0.85,"out":1.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"relace/relace-search","name":"Relace: Relace Search","org":"Relace","slug":"relace-search","matchKey":"relacesearch","blurb":"The relace-search model uses 4-12 `view_file` and `grep` tools in parallel to explore a codebase and return relevant files to the user request.","created":1765213560,"cutoff":null,"hf":null,"ctx":256000,"maxOut":128000,"in":["text"],"out":["text"],"price":{"in":1,"out":3},"caps":{"tools":true,"json":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Relace","quant":"bf16","in":1,"out":3,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"runway/aleph-2","name":"Runway: Aleph 2.0","org":"Runway","slug":"aleph-2","matchKey":"aleph2","blurb":"Runway Aleph 2.0 is an in-context video editing model from Runway.","created":1785339484,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image","video"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Runway","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"runway/gen-4.5","name":"Runway: Gen-4.5","org":"Runway","slug":"gen-4.5","matchKey":"gen45","blurb":"Runway Gen-4.5 is a video generation model from Runway for text-to-video and image-to-video workflows.","created":1785339483,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Runway","in":0,"out":0}],"scores":{},"perf":{"n":1},"open":false,"rankable":true,"scored":0},{"id":"sakana/fugu-max","name":"Sakana: Fugu Max","org":"Sakana","slug":"fugu-max","matchKey":"fugumax","blurb":"Fugu Max is the cost-performance model in Sakana AI's Fugu family.","created":1789104771,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":2,"out":6,"webSearch":0.01},"caps":{"tools":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high"]},"moderated":false,"providers":[{"name":"Sakana AI","in":2,"out":6,"cacheRead":0.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.25,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sakana/fugu-ultra","name":"Sakana: Fugu Ultra","org":"Sakana","slug":"fugu-ultra","matchKey":"fuguultra","blurb":"Fugu Ultra is the higher-performance model in Sakana AI's Fugu family.","created":1782276303,"cutoff":null,"hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image"],"out":["text"],"price":{"in":5,"out":30,"webSearch":0.01},"caps":{"tools":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high"]},"moderated":false,"providers":[{"name":"Sakana AI","in":5,"out":30,"cacheRead":0.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.5,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sakana/fugu-ultra-v2","name":"Sakana: Fugu Ultra v2","org":"Sakana","slug":"fugu-ultra-v2","matchKey":"fuguultrav2","blurb":"Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family.","created":1789105383,"cutoff":"2026-08-28","hf":null,"ctx":1000000,"maxOut":128000,"in":["text","image","file"],"out":["text"],"price":{"in":5,"out":30,"webSearch":0.01},"caps":{"tools":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","xhigh","high"]},"moderated":false,"providers":[{"name":"Sakana AI","in":5,"out":30,"cacheRead":0.5,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.5,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sakana/sakana-namazu","name":"Sakana: Sakana Namazu","org":"Sakana","slug":"sakana-namazu","matchKey":"sakananamazu","blurb":"Sakana Namazu is a Japanese-specialized reasoning model from Sakana AI, based on Kimi K2.6 with additional training for Japanese language and business contexts.","created":1786410129,"cutoff":null,"hf":null,"ctx":262144,"maxOut":65536,"in":["text","image","file"],"out":["text"],"price":{"in":0.95,"out":4,"webSearch":0.007},"caps":{"tools":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":["high","none"]},"moderated":false,"providers":[{"name":"Sakana AI","in":0.95,"out":4,"cacheRead":0.15,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.15,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sao10k/l3.3-euryale-70b","name":"Sao10K: Llama 3.3 Euryale 70B","org":"Sao10K","slug":"l3.3-euryale-70b","matchKey":"l33euryale70b","blurb":"Euryale L3.3 70B is a model focused on creative roleplay from [Sao10k](https://ko-fi.com/sao10k).","created":1734535928,"cutoff":"2023-12-31","hf":"Sao10K/L3.3-70B-Euryale-v2.3","ctx":131072,"maxOut":16384,"in":["text"],"out":["text"],"price":{"in":0.65,"out":0.75},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"NextBit","quant":"bf16","in":0.65,"out":0.75,"uptime":92.7}],"scores":{},"perf":{"uptime":92.7,"n":1},"open":true,"rankable":true,"scored":0},{"id":"sourceful/riverflow-v2-fast","name":"Sourceful: Riverflow V2 Fast","org":"Sourceful","slug":"riverflow-v2-fast","matchKey":"riverflowv2fast","blurb":"Riverflow V2 Fast is the fastest variant of Sourceful's Riverflow 2.0 lineup, best for production deployments and latency-critical workflows.","created":1770051423,"cutoff":null,"hf":null,"ctx":8192,"maxOut":7372,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000479041916167665},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Sourceful","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sourceful/riverflow-v2-pro","name":"Sourceful: Riverflow V2 Pro","org":"Sourceful","slug":"riverflow-v2-pro","matchKey":"riverflowv2pro","blurb":"Riverflow V2 Pro is the most powerful variant of Sourceful's Riverflow 2.0 lineup, best for top-tier control and perfect text rendering.","created":1770051427,"cutoff":null,"hf":null,"ctx":8192,"maxOut":7372,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000359281437125749},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Sourceful","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sourceful/riverflow-v2.5-fast","name":"Sourceful: Riverflow V2.5 Fast","org":"Sourceful","slug":"riverflow-v2.5-fast","matchKey":"riverflowv25fast","blurb":"Riverflow V2.5 Fast is the speed-optimized variant of Sourceful's Riverflow 2.5 lineup, best for production deployments and latency-critical workflows.","created":1780584983,"cutoff":null,"hf":null,"ctx":32768,"maxOut":29491,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.00000455089820359281},"caps":{"reasoning":true},"reasoning":{"mandatory":true,"efforts":["high","medium","low"]},"moderated":false,"providers":[{"name":"Sourceful","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"sourceful/riverflow-v2.5-pro","name":"Sourceful: Riverflow V2.5 Pro","org":"Sourceful","slug":"riverflow-v2.5-pro","matchKey":"riverflowv25pro","blurb":"Riverflow V2.5 Pro is the most powerful variant of Sourceful's Riverflow 2.5 lineup, best for top-tier control and quality-sensitive outputs.","created":1780584991,"cutoff":null,"hf":null,"ctx":32768,"maxOut":29491,"in":["text","image"],"out":["image"],"price":{"in":0,"out":0,"imageOut":0.0000311377245508982},"caps":{"reasoning":true},"reasoning":{"mandatory":true,"efforts":["xhigh","high","medium","low"]},"moderated":false,"providers":[{"name":"Sourceful","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"tencent/hunyuan-a13b-instruct","name":"Tencent: Hunyuan A13B Instruct","org":"Tencent","slug":"hunyuan-a13b-instruct","matchKey":"hunyuana13binstruct","blurb":"Hunyuan-A13B is a 13B active parameter Mixture-of-Experts (MoE) language model developed by Tencent, with a total parameter count of 80B and support for reasoning via Chain-of-Thought.","created":1751987664,"cutoff":"2025-03-31","hf":"tencent/Hunyuan-A13B-Instruct","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.14,"out":0.57},"caps":{"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"SiliconFlow","quant":"fp8","in":0.14,"out":0.57,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"tencent/hy-mt2-1.8b","name":"Tencent: Hy-MT2-1.8B","org":"Tencent","slug":"hy-mt2-1.8b","matchKey":"hymt218b","blurb":"Hy-MT2-1.8B is a compact 1.8B-parameter translation model from Tencent.","created":1787231581,"cutoff":null,"hf":"tencent/Hy-MT2-1.8B","ctx":8192,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":0.044,"out":0.177},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"Tencent","quant":"fp8","in":0.044,"out":0.177,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"tencent/hy-mt2-30b-a3b","name":"Tencent: Hy-MT2-30B-A3B","org":"Tencent","slug":"hy-mt2-30b-a3b","matchKey":"hymt230ba3b","blurb":"Hy-MT2-30B-A3B is Tencent's flagship translation model in the Hy-MT2 family.","created":1787231561,"cutoff":null,"hf":"tencent/Hy-MT2-30B-A3B","ctx":8192,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":0.074,"out":0.295},"caps":{"json":true,"schema":true},"reasoning":null,"moderated":false,"providers":[{"name":"Tencent","quant":"fp8","in":0.074,"out":0.295,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"tencent/hy-mt2-7b","name":"Tencent: Hy-MT2-7B","org":"Tencent","slug":"hy-mt2-7b","matchKey":"hymt27b","blurb":"Hy-MT2-7B is a 7B-parameter translation model from Tencent.","created":1787148797,"cutoff":null,"hf":"tencent/Hy-MT2-7B","ctx":8192,"maxOut":4096,"in":["text"],"out":["text"],"price":{"in":0.074,"out":0.295},"caps":{"json":true,"schema":true},"reasoning":null,"moderated":false,"providers":[{"name":"Tencent","quant":"fp8","in":0.074,"out":0.295,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"tencent/hy3-preview","name":"Tencent: Hy3 preview","org":"Tencent","slug":"hy3-preview","matchKey":"hy3preview","blurb":"Hy3 preview is a high-efficiency Mixture-of-Experts model from Tencent designed for agentic workflows and production use.","created":1776878150,"cutoff":null,"hf":"tencent/Hy3-preview","ctx":262144,"maxOut":235929,"in":["text"],"out":["text"],"price":{"in":0.18,"out":0.6},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":["high","low","none"]},"moderated":false,"providers":[{"name":"GMICloud","quant":"bf16","in":0.18,"out":0.6,"cacheRead":0.06,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.06,"n":1},"open":true,"rankable":true,"scored":0},{"id":"thedrummer/cydonia-24b-v4.1","name":"TheDrummer: Cydonia 24B V4.1","org":"TheDrummer","slug":"cydonia-24b-v4.1","matchKey":"cydonia24bv41","blurb":"Uncensored and creative writing model based on Mistral Small 3.2 24B with good recall, prompt adherence, and intelligence.","created":1758931878,"cutoff":"2024-04-30","hf":"thedrummer/cydonia-24b-v4.1","ctx":131072,"maxOut":117964,"in":["text"],"out":["text"],"price":{"in":0.3,"out":0.5},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"bf16","in":0.3,"out":0.5,"cacheRead":0.15,"uptime":99.9}],"scores":{},"perf":{"uptime":99.9,"cache":0.15,"n":1},"open":true,"rankable":true,"scored":0},{"id":"thedrummer/skyfall-36b-v2","name":"TheDrummer: Skyfall 36B V2","org":"TheDrummer","slug":"skyfall-36b-v2","matchKey":"skyfall36bv2","blurb":"Skyfall 36B v2 is an enhanced iteration of Mistral Small 2501, specifically fine-tuned for improved creativity, nuanced writing, role-playing, and coherent storytelling.","created":1741636566,"cutoff":"2024-06-30","hf":"TheDrummer/Skyfall-36B-v2","ctx":32768,"maxOut":29491,"in":["text"],"out":["text"],"price":{"in":0.55,"out":0.8},"caps":{"json":true,"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"fp8","in":0.55,"out":0.8,"cacheRead":0.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.25,"n":1},"open":true,"rankable":true,"scored":0},{"id":"thedrummer/unslopnemo-12b","name":"TheDrummer: UnslopNemo 12B","org":"TheDrummer","slug":"unslopnemo-12b","matchKey":"unslopnemo12b","blurb":"UnslopNemo v4.1 is the latest addition from the creator of Rocinante, designed for adventure writing and role-play scenarios.","created":1731103448,"cutoff":"2024-04-30","hf":"TheDrummer/UnslopNemo-12B-v4.1","ctx":1024000,"maxOut":819200,"in":["text"],"out":["text"],"price":{"in":0.4,"out":0.4},"caps":{"schema":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"Parasail","quant":"bf16","in":0.4,"out":0.4,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":true,"rankable":true,"scored":0},{"id":"typesafe/jev-1.13","name":"TypeSafe: Jev 1.13","org":"TypeSafe","slug":"jev-1.13","matchKey":"jev113","blurb":"Jev is a structured decision model from TypeSafe, and the first of its System One models.","created":1789689684,"cutoff":null,"hf":null,"ctx":32000,"maxOut":28800,"in":["text"],"out":["decisions"],"price":{"in":0.042,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"TypeSafe","in":0.042,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"unbiased/pareto","name":"Pareto","org":"unbiased","slug":"pareto","matchKey":"pareto","blurb":"Pareto is a multimodal composite model built for research, coding, and agentic workflows, while delivering frontier-level performance across a broad range of general-purpose tasks.","created":1789686178,"cutoff":null,"hf":null,"ctx":262144,"maxOut":131072,"in":["text","image"],"out":["text"],"price":{"in":2.5,"out":7.5},"caps":{"tools":true},"reasoning":null,"moderated":false,"providers":[{"name":"Unbiased","in":2.5,"out":7.5,"cacheRead":0.25,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.25,"n":1},"open":false,"rankable":true,"scored":0},{"id":"writer/palmyra-x5","name":"Writer: Palmyra X5","org":"Writer","slug":"palmyra-x5","matchKey":"palmyrax5","blurb":"Palmyra X5 is Writer's most advanced model, purpose-built for building and scaling AI agents across the enterprise.","created":1769003823,"cutoff":null,"hf":null,"ctx":1040000,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":0.6,"out":6},"caps":{},"reasoning":null,"moderated":true,"providers":[{"name":"Amazon Bedrock","in":0.6,"out":6,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"x-ai/grok-imagine-video","name":"SpaceXAI: Grok Imagine Video","org":"SpaceXAI","slug":"grok-imagine-video","matchKey":"grokimaginevideo","blurb":"Grok Imagine Video is SpaceXAI's fast, text-, image-, and reference-conditioned video generation model.","created":1779117586,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"xAI","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"x-ai/grok-imagine-video-1.5","name":"SpaceXAI: Grok Imagine Video 1.5","org":"SpaceXAI","slug":"grok-imagine-video-1.5","matchKey":"grokimaginevideo15","blurb":"Grok Imagine Video 1.5 is a video generation model from SpaceXAI.","created":1784548100,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["text","image"],"out":["video"],"price":{"in":0,"out":0},"caps":{"json":true,"logprobs":true,"seed":true},"reasoning":null,"moderated":false,"providers":[{"name":"xAI","in":0,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"x-ai/grok-stt-1.0","name":"SpaceXAI: Grok STT 1.0","org":"SpaceXAI","slug":"grok-stt-1.0","matchKey":"grokstt10","blurb":"Grok STT is SpaceXAI's speech-to-text model, available via the REST /v1/stt endpoint.","created":1784817014,"cutoff":null,"hf":null,"ctx":null,"maxOut":null,"in":["audio"],"out":["transcription"],"price":{"in":27.777778,"out":0},"caps":{},"reasoning":null,"moderated":false,"providers":[{"name":"xAI","in":27.777778,"out":0,"uptime":100}],"scores":{},"perf":{"uptime":100,"n":1},"open":false,"rankable":true,"scored":0},{"id":"xiaomi/mimo-v2.6-flash","name":"Xiaomi: MiMo-V2.6-Flash","org":"Xiaomi","slug":"mimo-v2.6-flash","matchKey":"mimov26flash","blurb":"MiMo-V2.6-Flash is an open-source foundation model developed by Xiaomi.","created":1790021264,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":131072,"in":["text","image","video","audio"],"out":["text"],"price":{"in":0.14,"out":0.28},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Xiaomi","quant":"fp8","in":0.14,"out":0.28,"cacheRead":0.0028,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.0028,"n":1},"open":false,"rankable":true,"scored":0},{"id":"xiaomi/mimo-v2.6-pro-ultraspeed","name":"Xiaomi: MiMo-V2.6-Pro-UltraSpeed","org":"Xiaomi","slug":"mimo-v2.6-pro-ultraspeed","matchKey":"mimov26proultraspeed","blurb":"MiMo-V2.6-Pro-UltraSpeed is the fast speed edition of Xiaomi's flagship foundation model, MiMo-V2.6-Pro.","created":1790021267,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":131072,"in":["text","image","video","audio"],"out":["text"],"price":{"in":4.35,"out":8.7},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Xiaomi","quant":"fp8","in":4.35,"out":8.7,"cacheRead":0.036,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.036,"n":1},"open":false,"rankable":true,"scored":0},{"id":"z-ai/glm-5-turbo","name":"Z.ai: GLM 5 Turbo","org":"Z.ai","slug":"glm-5-turbo","matchKey":"glm5turbo","blurb":"GLM-5 Turbo is a new model from Z.ai designed for fast inference and strong performance in agent-driven environments such as OpenClaw scenarios.","created":1773583573,"cutoff":null,"hf":null,"ctx":202752,"maxOut":131072,"in":["text"],"out":["text"],"price":{"in":1.2,"out":4},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[{"name":"Z.AI","in":1.2,"out":4,"cacheRead":0.24,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.24,"n":1},"open":false,"rankable":true,"scored":0},{"id":"z-ai/glm-5.3-flashx","name":"Z.ai: GLM 5.3 FlashX","org":"Z.ai","slug":"glm-5.3-flashx","matchKey":"glm53flashx","blurb":"GLM-5.3-FlashX is the high-speed variant of Z.ai's GLM-5.3-Flash, a native multimodal model delivering inference speeds of up to 200 tokens/s.","created":1789744020,"cutoff":null,"hf":null,"ctx":1048576,"maxOut":131072,"in":["text","image","video"],"out":["text"],"price":{"in":0.37,"out":1.25},"caps":{"tools":true,"json":true,"reasoning":true},"reasoning":{"mandatory":true,"efforts":["max","high","low"]},"moderated":false,"providers":[{"name":"Z.AI","quant":"fp8","in":0.37,"out":1.25,"cacheRead":0.075,"uptime":100}],"scores":{},"perf":{"uptime":100,"cache":0.075,"n":1},"open":false,"rankable":true,"scored":0},{"id":"cohere/north-mini-code","name":"Cohere: North Mini Code (free)","org":"Cohere","slug":"north-mini-code","matchKey":"northminicode","blurb":"North Mini Code is Cohere's first agentic coding model and the debut of its North family.","created":1781723748,"cutoff":null,"hf":"CohereLabs/North-Mini-Code-1.0","ctx":256000,"maxOut":64000,"in":["text"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":true,"providers":[],"scores":{},"perf":{"n":0},"open":true,"rankable":false,"scored":0},{"id":"dots-studio/dots-3-note-preview","name":"Dots Studio: Dots3-Note Preview (free)","org":"Dots Studio","slug":"dots-3-note-preview","matchKey":"dots3notepreview","blurb":"Dots3-Note Preview is an open-weight mixture-of-experts model from Dots Studio, with 16B active parameters out of 280B total.","created":1786680361,"cutoff":null,"hf":null,"ctx":512000,"maxOut":460800,"in":["text","image"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":false,"rankable":false,"scored":0},{"id":"inclusionai/ling-3.0-flash-sante","name":"inclusionAI: Ling 3.0 Flash Sante (free)","org":"inclusionAI","slug":"ling-3.0-flash-sante","matchKey":"ling30flashsante","blurb":"Ling 3.0 Flash Sante is a health and medicine-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total.","created":1788545946,"cutoff":null,"hf":null,"ctx":262144,"maxOut":32768,"in":["text"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":false,"rankable":false,"scored":0},{"id":"liquid/lfm-2.5-2.6b","name":"LiquidAI: LFM2.5-2.6B (free)","org":"LiquidAI","slug":"lfm-2.5-2.6b","matchKey":"lfm2526b","blurb":"LFM2.5-2.6B is a compact reasoning model from Liquid AI.","created":1786470519,"cutoff":null,"hf":"LiquidAI/LFM2.5-2.6B","ctx":65536,"maxOut":8192,"in":["text"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true,"seed":true},"reasoning":{"mandatory":true,"efforts":[]},"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":true,"rankable":false,"scored":0},{"id":"mistralai/mistral-large-2512","name":"Mistral: Mistral Large 3 2512 (batch)","org":"Mistral","slug":"mistral-large-2512","matchKey":"mistrallarge2512","blurb":"Mistral Large 3 2512 is Mistral’s most capable model to date, featuring a sparse mixture-of-experts architecture with 41B active parameters (675B total), and released under the Apache 2.0 license.","created":1764624472,"cutoff":null,"hf":null,"ctx":262144,"maxOut":209715,"in":["text","image","file"],"out":["text"],"price":{"in":0.25,"out":0.75,"alt":{"batch":{"in":0.25,"out":0.75}}},"caps":{"tools":true,"json":true,"schema":true,"seed":true},"reasoning":null,"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":false,"rankable":false,"scored":0},{"id":"nex-agi/nex-n2.5-mini","name":"Nex AGI: Nex-N2.5-Mini (free)","org":"Nex AGI","slug":"nex-n2.5-mini","matchKey":"nexn25mini","blurb":"Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes.","created":1788890061,"cutoff":null,"hf":"nex-agi/Nex-N2.5-mini","ctx":262144,"maxOut":235929,"in":["text","image"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true},"reasoning":{"mandatory":false,"efforts":["high","medium","none"]},"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":true,"rankable":false,"scored":0},{"id":"nex-agi/nex-n2.5-pro","name":"Nex AGI: Nex-N2.5-Pro (free)","org":"Nex AGI","slug":"nex-n2.5-pro","matchKey":"nexn25pro","blurb":"Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes.","created":1788890050,"cutoff":null,"hf":"nex-agi/Nex-N2.5-Pro","ctx":262144,"maxOut":235929,"in":["text","image"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"json":true,"schema":true,"reasoning":true,"logprobs":true},"reasoning":{"mandatory":false,"efforts":["high","medium","none"]},"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":true,"rankable":false,"scored":0},{"id":"nvidia/nemotron-3-nano-omni-30b-a3b-reasoning","name":"NVIDIA: Nemotron 3 Nano Omni (free)","org":"NVIDIA","slug":"nemotron-3-nano-omni-30b-a3b-reasoning","matchKey":"nemotron3nanoomni30ba3breasoning","blurb":"NVIDIA Nemotron™ 3 Nano Omni is a 30B-A3B open multimodal model designed to function as a perception and context sub-agent in enterprise agent systems.","created":1777393095,"cutoff":null,"hf":"nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16","ctx":256000,"maxOut":65536,"in":["text","audio","image","video"],"out":["text"],"price":{"in":0,"out":0,"alt":{"free":{"in":0,"out":0}}},"caps":{"tools":true,"reasoning":true,"seed":true},"reasoning":{"mandatory":false,"efforts":[]},"moderated":false,"providers":[],"scores":{},"perf":{"n":0},"open":true,"rankable":false,"scored":0}]}