AI models

How the labs evolved

Each lab's record per benchmark over time, and price per token when each model was listed on OpenRouter.

Record per lab: GPQA Diamond ?what it is

0%25%50%75%100%202420252026Google: Gemini 1.0 Pro, 34%, 2023Google: Gemini 1.5 Pro, 45.9%, 2024Google: Gemini 1.5 Flash, 47.3%, 2024Google: Gemini 1.5 Pro, 57.2%, 2024Google: Gemini 2.0 Flash, 64.1%, 2025Google: Gemini 2.0 Pro, 65.7%, 2025Google: Gemini 2.5 Pro (Mar 2025), 83.8%, 2025Google: Gemini 2.5 Pro (Jun 2025), 84.8%, 2025Google: Gemini 2.5 Pro (Jun 2025), 85.3%, 2025Google: Gemini 3 Pro, 92.6%, 2025Google: Gemini 3.1 Pro, 94.1%, 2026Google: Gemini 3.1 Pro, 94.4%, 2026Google: Gemini 3.7 Flash, 94.8%, 2026Google: Gemini 3.8 Flash, 95.4%, 2026Google: Gemini 4 Argon, 96.4%, 2026OpenAI: GPT-4 (Mar 2023), 35.7%, 2023OpenAI: GPT-4 Turbo (Nov 2023), 42.4%, 2023OpenAI: GPT-4 Turbo (Apr 2024), 46.6%, 2024OpenAI: GPT-4o, 48.9%, 2024OpenAI: GPT-4o, 49.2%, 2024OpenAI: o1-preview, 50.3%, 2024OpenAI: o1-mini, 59.5%, 2024OpenAI: o1-mini, 62.4%, 2024OpenAI: o1, 74.2%, 2024OpenAI: o1, 75.8%, 2024OpenAI: o1, 76.8%, 2024OpenAI: o3-mini, 77%, 2025OpenAI: o4-mini, 77.8%, 2025OpenAI: o4-mini, 79.6%, 2025OpenAI: o3, 79.8%, 2025OpenAI: o3, 80.8%, 2025OpenAI: o3, 81.8%, 2025OpenAI: GPT-5, 85.4%, 2025OpenAI: GPT-5, 86.2%, 2025OpenAI: GPT-5.1, 87.6%, 2025OpenAI: GPT-5.2, 87.9%, 2025OpenAI: GPT-5.2, 88.2%, 2025OpenAI: GPT-5.2, 91.4%, 2025OpenAI: GPT-5.4, 93.3%, 2026OpenAI: GPT-5.4 Pro, 94.6%, 2026OpenAI: GPT-6 Astra, 95.8%, 2026Anthropic: Claude 2, 34.7%, 2023Anthropic: Claude 3 Sonnet, 40.6%, 2024Anthropic: Claude 3 Opus, 47.2%, 2024Anthropic: Claude 3.5 Sonnet, 54%, 2024Anthropic: Claude 3.5 Sonnet (October 2024), 55.3%, 2024Anthropic: Claude 3.7 Sonnet, 66%, 2025Anthropic: Claude 3.7 Sonnet, 76.8%, 2025Anthropic: Claude 3.7 Sonnet, 79.7%, 2025Anthropic: Claude Sonnet 4.5, 81.7%, 2025Anthropic: Claude Sonnet 4.5, 82.3%, 2025Anthropic: Claude Opus 4.5, 85.5%, 2025Anthropic: Claude Opus 4.5, 86%, 2025Anthropic: Claude Opus 4.6, 88.4%, 2026Anthropic: Claude Opus 4.6, 88.8%, 2026Anthropic: Claude Opus 4.6, 90.5%, 2026Anthropic: Claude Opus 4.8, 91%, 2026Anthropic: Claude Opus 5, 92.9%, 2026Anthropic: Claude Opus 5, 93.9%, 2026Anthropic: Claude Sonnet 5.5, 95.6%, 2026Moonshot AI: Kimi K2 Thinking, 84.2%, 2025Moonshot AI: Kimi K2.5, 87.6%, 2026Moonshot AI: Kimi K2.6, 90.8%, 2026Moonshot AI: Kimi K3, 91.9%, 2026Moonshot AI: Kimi K3, 93.1%, 2026Moonshot AI: Kimi K3.1 Pro, 94.2%, 2026xAI: Grok-2, 53.8%, 2024xAI: Grok-3 mini, 75.5%, 2025xAI: Grok 3, 75.8%, 2025xAI: Grok-3 mini, 76.3%, 2025xAI: Grok 4, 87%, 2025xAI: Grok 4.20, 89.3%, 2026xAI: Grok 4.5, 93.4%, 2026xAI: Grok 4.6, 94%, 2026DeepSeek: DeepSeek LLM 67B, 24.6%, 2023DeepSeek: DeepSeek-V3, 56.5%, 2024DeepSeek: DeepSeek-R1-Distill-Qwen-32B, 64.1%, 2025DeepSeek: DeepSeek-R1, 71.7%, 2025DeepSeek: DeepSeek-R1 (May 2025), 76.3%, 2025DeepSeek: DeepSeek-V3.2, 83.4%, 2025DeepSeek: DeepSeek-V4-Pro, 89.6%, 2026DeepSeek: DeepSeek-V4-Pro, 90.9%, 2026DeepSeek: DeepSeek V4 Flash 0731, 91%, 2026DeepSeek: DeepSeek V4 Pro 0813, 91.7%, 2026DeepSeek: DeepSeek V4.1 Flash 1017, 93.2%, 2026ByteDance: Doubao-Seed-1.6, 77.6%, 2025ByteDance: Doubao-Seed-2.0 Pro, 88.2%, 2026ByteDance: Doubao Seed 2.1 Pro, 89.3%, 2026ByteDance: Doubao Seed 2.1 Turbo, 89.6%, 2026ByteDance: Doubao Seed Evolving (Nov 2026), 93.1%, 2026Alibaba: Qwen1.5-72B, 28.8%, 2024Alibaba: Qwen1.5-32B, 30.7%, 2024Alibaba: Qwen2-72B, 40.8%, 2024Alibaba: Qwen2.5-32B, 46.1%, 2024Alibaba: Qwen2.5-72B, 49.1%, 2024Alibaba: Qwen2.5-Max, 56.1%, 2025Alibaba: QwQ-32B, 65.3%, 2025Alibaba: QWQ-Plus, 65.4%, 2025Alibaba: Qwen3-32B, 65.7%, 2025Alibaba: Qwen3-235B-A22B, 70.7%, 2025Alibaba: Qwen3-235B-A22B (Jul 2025), 80.1%, 2025Alibaba: Qwen3.5 397B-A17B, 85.9%, 2026Alibaba: Qwen3.5 397B-A17B, 86.4%, 2026Alibaba: Qwen 3.6 Plus, 88.4%, 2026Alibaba: Qwen3.7-Max, 90.9%, 2026Alibaba: Qwen 3.8 Max, 92.7%, 2026Zhipu AI (Z.ai): GLM-4.7, 83.3%, 2025Zhipu AI (Z.ai): GLM-5, 87.8%, 2026Zhipu AI (Z.ai): GLM-5.1, 89.9%, 2026Zhipu AI (Z.ai): GLM-5.2, 91.9%, 2026MiniMax: MiniMax-M3, 81.3%, 2026MiniMax: MiniMax-M3, 90.9%, 2026Mistral AI: Mistral 7B, 13.2%, 2023Mistral AI: Mixtral 8x7B, 29.8%, 2023Mistral AI: Mixtral 8x7B, 30.6%, 2023Mistral AI: Mistral Large, 38.8%, 2024Mistral AI: Mistral Large 2, 49%, 2024Mistral AI: Mistral Large 2, 51.3%, 2024Mistral AI: Mistral Medium 3, 59.5%, 2025Mistral AI: Mistral Large 4, 88.1%, 2026Mistral AI: Mistral Large 4, 90.2%, 2026Meta: Llama 2-70B, 26.3%, 2023Meta: Llama 3-70B, 40.6%, 2024Meta: Llama 3.1-70B, 44.2%, 2024Meta: Llama 3.1-405B, 50.9%, 2024Meta: Llama 4 Scout, 51.8%, 2025Meta: Llama 4 Maverick, 67%, 2025Meta: Muse Spark, 89.8%, 2026NVIDIA: Nemotron 3 Ultra, 85.4%, 2026Microsoft: WizardLM-2 8x22B, 43.4%, 2024Microsoft: Phi-4, 56.1%, 2024Ai2 (Allen Institute): Tulu 3 (Tülu 3) 70B, 46.3%, 202401.AI: Yi-34B, 14.7%, 202301.AI: Yi-1.5-34B, 32%, 2024
  • Google 96.4%
  • OpenAI 95.8%
  • Anthropic 95.6%
  • Moonshot AI 94.2%
  • xAI 94%
  • DeepSeek 93.2%
  • ByteDance 93.1%
  • Alibaba 92.7%
  • Zhipu AI (Z.ai) 91.9%
  • MiniMax 90.9%
  • Mistral AI 90.2%
  • Meta 89.8%
  • NVIDIA 85.4%
  • Microsoft 56.1%
  • Ai2 (Allen Institute) 46.3%
  • 01.AI 32%

Source: Epoch AI, CC BY 4.0, latest version we checked. The 90% to 100% band marks the saturation zone.

Chart data table
LabModel%Dates in sources
AnthropicClaude 234.7%2023
AnthropicClaude 3 Sonnet40.6%2024
AnthropicClaude 3 Opus47.2%2024
AnthropicClaude 3.5 Sonnet54%2024
AnthropicClaude 3.5 Sonnet (October 2024)55.3%2024
AnthropicClaude 3.7 Sonnet66%2025
AnthropicClaude 3.7 Sonnet76.8%2025
AnthropicClaude 3.7 Sonnet79.7%2025
AnthropicClaude Sonnet 4.581.7%2025
AnthropicClaude Sonnet 4.582.3%2025
AnthropicClaude Opus 4.585.5%2025
AnthropicClaude Opus 4.586%2025
AnthropicClaude Opus 4.688.4%2026
AnthropicClaude Opus 4.688.8%2026
AnthropicClaude Opus 4.690.5%2026
AnthropicClaude Opus 4.891%2026
AnthropicClaude Opus 592.9%2026
AnthropicClaude Opus 593.9%2026
AnthropicClaude Sonnet 5.595.6%2026
Moonshot AIKimi K2 Thinking84.2%2025
Moonshot AIKimi K2.587.6%2026
Moonshot AIKimi K2.690.8%2026
Moonshot AIKimi K391.9%2026
Moonshot AIKimi K393.1%2026
Moonshot AIKimi K3.1 Pro94.2%2026
ByteDanceDoubao-Seed-1.677.6%2025
ByteDanceDoubao-Seed-2.0 Pro88.2%2026
ByteDanceDoubao Seed 2.1 Pro89.3%2026
ByteDanceDoubao Seed 2.1 Turbo89.6%2026
ByteDanceDoubao Seed Evolving (Nov 2026)93.1%2026
Mistral AIMistral 7B13.2%2023
Mistral AIMixtral 8x7B29.8%2023
Mistral AIMixtral 8x7B30.6%2023
Mistral AIMistral Large38.8%2024
Mistral AIMistral Large 249%2024
Mistral AIMistral Large 251.3%2024
Mistral AIMistral Medium 359.5%2025
Mistral AIMistral Large 488.1%2026
Mistral AIMistral Large 490.2%2026
OpenAIGPT-4 (Mar 2023)35.7%2023
OpenAIGPT-4 Turbo (Nov 2023)42.4%2023
OpenAIGPT-4 Turbo (Apr 2024)46.6%2024
OpenAIGPT-4o48.9%2024
OpenAIGPT-4o49.2%2024
OpenAIo1-preview50.3%2024
OpenAIo1-mini59.5%2024
OpenAIo1-mini62.4%2024
OpenAIo174.2%2024
OpenAIo175.8%2024
OpenAIo176.8%2024
OpenAIo3-mini77%2025
OpenAIo4-mini77.8%2025
OpenAIo4-mini79.6%2025
OpenAIo379.8%2025
OpenAIo380.8%2025
OpenAIo381.8%2025
OpenAIGPT-585.4%2025
OpenAIGPT-586.2%2025
OpenAIGPT-5.187.6%2025
OpenAIGPT-5.287.9%2025
OpenAIGPT-5.288.2%2025
OpenAIGPT-5.291.4%2025
OpenAIGPT-5.493.3%2026
OpenAIGPT-5.4 Pro94.6%2026
OpenAIGPT-6 Astra95.8%2026
DeepSeekDeepSeek LLM 67B24.6%2023
DeepSeekDeepSeek-V356.5%2024
DeepSeekDeepSeek-R1-Distill-Qwen-32B64.1%2025
DeepSeekDeepSeek-R171.7%2025
DeepSeekDeepSeek-R1 (May 2025)76.3%2025
DeepSeekDeepSeek-V3.283.4%2025
DeepSeekDeepSeek-V4-Pro89.6%2026
DeepSeekDeepSeek-V4-Pro90.9%2026
DeepSeekDeepSeek V4 Flash 073191%2026
DeepSeekDeepSeek V4 Pro 081391.7%2026
DeepSeekDeepSeek V4.1 Flash 101793.2%2026
MetaLlama 2-70B26.3%2023
MetaLlama 3-70B40.6%2024
MetaLlama 3.1-70B44.2%2024
MetaLlama 3.1-405B50.9%2024
MetaLlama 4 Scout51.8%2025
MetaLlama 4 Maverick67%2025
MetaMuse Spark89.8%2026
GoogleGemini 1.0 Pro34%2023
GoogleGemini 1.5 Pro45.9%2024
GoogleGemini 1.5 Flash47.3%2024
GoogleGemini 1.5 Pro57.2%2024
GoogleGemini 2.0 Flash64.1%2025
GoogleGemini 2.0 Pro65.7%2025
GoogleGemini 2.5 Pro (Mar 2025)83.8%2025
GoogleGemini 2.5 Pro (Jun 2025)84.8%2025
GoogleGemini 2.5 Pro (Jun 2025)85.3%2025
GoogleGemini 3 Pro92.6%2025
GoogleGemini 3.1 Pro94.1%2026
GoogleGemini 3.1 Pro94.4%2026
GoogleGemini 3.7 Flash94.8%2026
GoogleGemini 3.8 Flash95.4%2026
GoogleGemini 4 Argon96.4%2026
xAIGrok-253.8%2024
xAIGrok-3 mini75.5%2025
xAIGrok 375.8%2025
xAIGrok-3 mini76.3%2025
xAIGrok 487%2025
xAIGrok 4.2089.3%2026
xAIGrok 4.593.4%2026
xAIGrok 4.694%2026
AlibabaQwen1.5-72B28.8%2024
AlibabaQwen1.5-32B30.7%2024
AlibabaQwen2-72B40.8%2024
AlibabaQwen2.5-32B46.1%2024
AlibabaQwen2.5-72B49.1%2024
AlibabaQwen2.5-Max56.1%2025
AlibabaQwQ-32B65.3%2025
AlibabaQWQ-Plus65.4%2025
AlibabaQwen3-32B65.7%2025
AlibabaQwen3-235B-A22B70.7%2025
AlibabaQwen3-235B-A22B (Jul 2025)80.1%2025
AlibabaQwen3.5 397B-A17B85.9%2026
AlibabaQwen3.5 397B-A17B86.4%2026
AlibabaQwen 3.6 Plus88.4%2026
AlibabaQwen3.7-Max90.9%2026
AlibabaQwen 3.8 Max92.7%2026
Zhipu AI (Z.ai)GLM-4.783.3%2025
Zhipu AI (Z.ai)GLM-587.8%2026
Zhipu AI (Z.ai)GLM-5.189.9%2026
Zhipu AI (Z.ai)GLM-5.291.9%2026
NVIDIANemotron 3 Ultra85.4%2026
MiniMaxMiniMax-M381.3%2026
MiniMaxMiniMax-M390.9%2026
MicrosoftWizardLM-2 8x22B43.4%2024
MicrosoftPhi-456.1%2024
01.AIYi-34B14.7%2023
01.AIYi-1.5-34B32%2024
Ai2 (Allen Institute)Tulu 3 (Tülu 3) 70B46.3%2024

Record per lab: OTIS Mock AIME ?

0%25%50%75%100%202420252026Anthropic: Claude 2, 2.5%, 2023Anthropic: Claude 3 Opus, 4.7%, 2024Anthropic: Claude 3.5 Sonnet, 6.5%, 2024Anthropic: Claude 3.5 Sonnet (October 2024), 8.5%, 2024Anthropic: Claude 3.7 Sonnet, 21.9%, 2025Anthropic: Claude 3.7 Sonnet, 46.7%, 2025Anthropic: Claude 3.7 Sonnet, 53.3%, 2025Anthropic: Claude 3.7 Sonnet, 57.8%, 2025Anthropic: Claude Opus 4, 60%, 2025Anthropic: Claude Opus 4, 64.4%, 2025Anthropic: Claude Sonnet 4, 68.9%, 2025Anthropic: Claude Sonnet 4, 71.1%, 2025Anthropic: Claude Sonnet 4.5, 77.8%, 2025Anthropic: Claude Opus 4.5, 81.7%, 2025Anthropic: Claude Opus 4.5, 86.1%, 2025Anthropic: Claude Opus 4.6, 91.1%, 2026Anthropic: Claude Opus 4.6, 93.1%, 2026Anthropic: Claude Opus 4.6, 94.4%, 2026Anthropic: Claude Opus 4.7, 97.8%, 2026Anthropic: Claude Opus 4.8, 98.3%, 2026Anthropic: Claude Fable 5, 99.7%, 2026Anthropic: Claude Fable 5, 100%, 2026Moonshot AI: Kimi K2 Thinking, 83.1%, 2025Moonshot AI: Kimi K2.5, 92.2%, 2026Moonshot AI: Kimi K2.6, 96.1%, 2026Moonshot AI: Kimi K3, 97.2%, 2026Moonshot AI: Kimi K3.1 Pro, 100%, 2026OpenAI: GPT-4 (Mar 2023), 0.6%, 2023OpenAI: GPT-4 (Jun 2023), 1.1%, 2023OpenAI: GPT-3.5 Turbo, 2.2%, 2024OpenAI: GPT-4 Turbo (Apr 2024), 6.7%, 2024OpenAI: GPT-4o mini, 6.9%, 2024OpenAI: o1-preview, 31.1%, 2024OpenAI: o1-mini, 44.7%, 2024OpenAI: o1-mini, 46.9%, 2024OpenAI: o1, 53.3%, 2024OpenAI: o1, 73.3%, 2024OpenAI: o3-mini, 76.9%, 2025OpenAI: o4-mini, 81.7%, 2025OpenAI: o3, 83.9%, 2025OpenAI: o3, 84.4%, 2025OpenAI: gpt-oss-120b, 88.9%, 2025OpenAI: GPT-5, 91.4%, 2025OpenAI: GPT-5.2, 93.9%, 2025OpenAI: GPT-5.2, 96.1%, 2025OpenAI: GPT-5.2, 96.1%, 2025OpenAI: GPT-5.4, 97.8%, 2026OpenAI: GPT-5.5, 100%, 2026Google: Gemini 1.0 Pro, 1.1%, 2023Google: Gemini 1.5 Pro, 6.8%, 2024Google: Gemini 1.5 Flash, 16.2%, 2024Google: Gemini 1.5 Pro, 23.1%, 2024Google: Gemini 2.0 Flash Thinking, 57.8%, 2025Google: Gemini 2.5 Flash, 73.1%, 2025Google: Gemini 2.5 Pro (Jun 2025), 84.7%, 2025Google: Gemini 3 Pro, 91.4%, 2025Google: Gemini 3 Flash, 92.8%, 2025Google: Gemini 3 Flash, 95.6%, 2025Google: Gemini 3.1 Pro, 95.6%, 2026Google: Gemini 3.7 Flash, 97.2%, 2026Google: Gemini 3.8 Flash, 98.9%, 2026Google: Gemini 4 Argon, 100%, 2026Alibaba: Qwen2.5-32B, 7.4%, 2024Alibaba: Qwen2.5-72B, 8.1%, 2024Alibaba: Qwen2.5-Max, 16.1%, 2025Alibaba: Qwen Plus, 17.8%, 2025Alibaba: QwQ-32B, 59.2%, 2025Alibaba: Qwen3-30B-A3B, 62.8%, 2025Alibaba: Qwen3-14B, 66.4%, 2025Alibaba: Qwen3-32B, 66.9%, 2025Alibaba: Qwen3-235B-A22B (Jul 2025), 86.7%, 2025Alibaba: Qwen3.5 397B-A17B, 88.9%, 2026Alibaba: Qwen 3.6 Plus, 93.3%, 2026Alibaba: Qwen3.7-Max, 95.6%, 2026Alibaba: Qwen 3.8 Max, 99.4%, 2026Alibaba: Qwen3.8 Max (0902), 100%, 2026Mistral AI: Mistral Large, 1.9%, 2024Mistral AI: Mistral Large 2, 8.5%, 2024Mistral AI: Mistral Medium 3, 32.2%, 2025Mistral AI: Mistral Large 4, 98.3%, 2026Mistral AI: Mistral Large 4, 99.4%, 2026DeepSeek: DeepSeek LLM 67B, 0.8%, 2023DeepSeek: DeepSeek-V3, 15.8%, 2024DeepSeek: DeepSeek-R1-Distill-Qwen-1.5B, 21.4%, 2025DeepSeek: DeepSeek-R1-Distill-Qwen-14B, 50.6%, 2025DeepSeek: DeepSeek-R1-Distill-Llama-70B, 51.4%, 2025DeepSeek: DeepSeek-R1, 53.3%, 2025DeepSeek: DeepSeek-R1-Distill-Qwen-32B, 55.6%, 2025DeepSeek: DeepSeek-R1 (May 2025), 66.4%, 2025DeepSeek: DeepSeek-V3.2, 87.8%, 2025DeepSeek: DeepSeek-V4-Pro, 95.6%, 2026DeepSeek: DeepSeek-V4-Pro, 96.7%, 2026DeepSeek: DeepSeek V4 Pro 0813, 98.6%, 2026DeepSeek: DeepSeek V4.1 Flash 1017, 99.4%, 2026Meta: Llama 2-70B, 0%, 2023Meta: Llama 3-8B, 1.9%, 2024Meta: Llama 3-70B, 4.3%, 2024Meta: Llama 3.1-405B, 9.7%, 2024Meta: Llama 4 Maverick, 20.6%, 2025Meta: Muse Spark, 88.9%, 2026Meta: Muse Spark 1.3, 99.2%, 2026Meta: Muse Wonder, 99.4%, 2026ByteDance: Doubao-Seed-1.6, 69.4%, 2025ByteDance: Doubao-Seed-2.0 Pro, 93.9%, 2026ByteDance: Doubao Seed 2.1 Pro, 95.6%, 2026ByteDance: Doubao Seed 2.1 Turbo, 96.1%, 2026ByteDance: Doubao Seed Evolving (Nov 2026), 99.2%, 2026xAI: Grok-2, 11.5%, 2024xAI: Grok 3, 55.6%, 2025xAI: Grok-3 mini, 62.2%, 2025xAI: Grok-3 mini, 77.8%, 2025xAI: Grok 4, 84%, 2025xAI: Grok 4.20, 92.2%, 2026xAI: Grok 4.3 Beta, 93.3%, 2026xAI: Grok 4.5, 97.8%, 2026xAI: Grok 4.6, 99.2%, 2026Zhipu AI (Z.ai): GLM-4.7, 83.3%, 2025Zhipu AI (Z.ai): GLM-5.1, 93.3%, 2026Zhipu AI (Z.ai): GLM-5.3-Flash, 93.9%, 2026NVIDIA: Nemotron 3 Ultra, 86.7%, 2026MiniMax: MiniMax-M3, 26.7%, 2026MiniMax: MiniMax-M3, 71.1%, 2026Microsoft: Phi-4, 13.8%, 2024Ai2 (Allen Institute): Tulu 3 (Tülu 3) 70B, 4.4%, 2024
  • Anthropic 100%
  • Moonshot AI 100%
  • OpenAI 100%
  • Google 100%
  • Alibaba 100%
  • Mistral AI 99.4%
  • DeepSeek 99.4%
  • Meta 99.4%
  • ByteDance 99.2%
  • xAI 99.2%
  • Zhipu AI (Z.ai) 93.9%
  • NVIDIA 86.7%
  • MiniMax 71.1%
  • Microsoft 13.8%
  • Ai2 (Allen Institute) 4.4%

Source: Epoch AI, CC BY 4.0, latest version we checked. The 90% to 100% band marks the saturation zone.

Chart data table
LabModel%Dates in sources
AnthropicClaude 22.5%2023
AnthropicClaude 3 Opus4.7%2024
AnthropicClaude 3.5 Sonnet6.5%2024
AnthropicClaude 3.5 Sonnet (October 2024)8.5%2024
AnthropicClaude 3.7 Sonnet21.9%2025
AnthropicClaude 3.7 Sonnet46.7%2025
AnthropicClaude 3.7 Sonnet53.3%2025
AnthropicClaude 3.7 Sonnet57.8%2025
AnthropicClaude Opus 460%2025
AnthropicClaude Opus 464.4%2025
AnthropicClaude Sonnet 468.9%2025
AnthropicClaude Sonnet 471.1%2025
AnthropicClaude Sonnet 4.577.8%2025
AnthropicClaude Opus 4.581.7%2025
AnthropicClaude Opus 4.586.1%2025
AnthropicClaude Opus 4.691.1%2026
AnthropicClaude Opus 4.693.1%2026
AnthropicClaude Opus 4.694.4%2026
AnthropicClaude Opus 4.797.8%2026
AnthropicClaude Opus 4.898.3%2026
AnthropicClaude Fable 599.7%2026
AnthropicClaude Fable 5100%2026
Moonshot AIKimi K2 Thinking83.1%2025
Moonshot AIKimi K2.592.2%2026
Moonshot AIKimi K2.696.1%2026
Moonshot AIKimi K397.2%2026
Moonshot AIKimi K3.1 Pro100%2026
ByteDanceDoubao-Seed-1.669.4%2025
ByteDanceDoubao-Seed-2.0 Pro93.9%2026
ByteDanceDoubao Seed 2.1 Pro95.6%2026
ByteDanceDoubao Seed 2.1 Turbo96.1%2026
ByteDanceDoubao Seed Evolving (Nov 2026)99.2%2026
Mistral AIMistral Large1.9%2024
Mistral AIMistral Large 28.5%2024
Mistral AIMistral Medium 332.2%2025
Mistral AIMistral Large 498.3%2026
Mistral AIMistral Large 499.4%2026
OpenAIGPT-4 (Mar 2023)0.6%2023
OpenAIGPT-4 (Jun 2023)1.1%2023
OpenAIGPT-3.5 Turbo2.2%2024
OpenAIGPT-4 Turbo (Apr 2024)6.7%2024
OpenAIGPT-4o mini6.9%2024
OpenAIo1-preview31.1%2024
OpenAIo1-mini44.7%2024
OpenAIo1-mini46.9%2024
OpenAIo153.3%2024
OpenAIo173.3%2024
OpenAIo3-mini76.9%2025
OpenAIo4-mini81.7%2025
OpenAIo383.9%2025
OpenAIo384.4%2025
OpenAIgpt-oss-120b88.9%2025
OpenAIGPT-591.4%2025
OpenAIGPT-5.293.9%2025
OpenAIGPT-5.296.1%2025
OpenAIGPT-5.296.1%2025
OpenAIGPT-5.497.8%2026
OpenAIGPT-5.5100%2026
DeepSeekDeepSeek LLM 67B0.8%2023
DeepSeekDeepSeek-V315.8%2024
DeepSeekDeepSeek-R1-Distill-Qwen-1.5B21.4%2025
DeepSeekDeepSeek-R1-Distill-Qwen-14B50.6%2025
DeepSeekDeepSeek-R1-Distill-Llama-70B51.4%2025
DeepSeekDeepSeek-R153.3%2025
DeepSeekDeepSeek-R1-Distill-Qwen-32B55.6%2025
DeepSeekDeepSeek-R1 (May 2025)66.4%2025
DeepSeekDeepSeek-V3.287.8%2025
DeepSeekDeepSeek-V4-Pro95.6%2026
DeepSeekDeepSeek-V4-Pro96.7%2026
DeepSeekDeepSeek V4 Pro 081398.6%2026
DeepSeekDeepSeek V4.1 Flash 101799.4%2026
MetaLlama 2-70B0%2023
MetaLlama 3-8B1.9%2024
MetaLlama 3-70B4.3%2024
MetaLlama 3.1-405B9.7%2024
MetaLlama 4 Maverick20.6%2025
MetaMuse Spark88.9%2026
MetaMuse Spark 1.399.2%2026
MetaMuse Wonder99.4%2026
GoogleGemini 1.0 Pro1.1%2023
GoogleGemini 1.5 Pro6.8%2024
GoogleGemini 1.5 Flash16.2%2024
GoogleGemini 1.5 Pro23.1%2024
GoogleGemini 2.0 Flash Thinking57.8%2025
GoogleGemini 2.5 Flash73.1%2025
GoogleGemini 2.5 Pro (Jun 2025)84.7%2025
GoogleGemini 3 Pro91.4%2025
GoogleGemini 3 Flash92.8%2025
GoogleGemini 3 Flash95.6%2025
GoogleGemini 3.1 Pro95.6%2026
GoogleGemini 3.7 Flash97.2%2026
GoogleGemini 3.8 Flash98.9%2026
GoogleGemini 4 Argon100%2026
xAIGrok-211.5%2024
xAIGrok 355.6%2025
xAIGrok-3 mini62.2%2025
xAIGrok-3 mini77.8%2025
xAIGrok 484%2025
xAIGrok 4.2092.2%2026
xAIGrok 4.3 Beta93.3%2026
xAIGrok 4.597.8%2026
xAIGrok 4.699.2%2026
AlibabaQwen2.5-32B7.4%2024
AlibabaQwen2.5-72B8.1%2024
AlibabaQwen2.5-Max16.1%2025
AlibabaQwen Plus17.8%2025
AlibabaQwQ-32B59.2%2025
AlibabaQwen3-30B-A3B62.8%2025
AlibabaQwen3-14B66.4%2025
AlibabaQwen3-32B66.9%2025
AlibabaQwen3-235B-A22B (Jul 2025)86.7%2025
AlibabaQwen3.5 397B-A17B88.9%2026
AlibabaQwen 3.6 Plus93.3%2026
AlibabaQwen3.7-Max95.6%2026
AlibabaQwen 3.8 Max99.4%2026
AlibabaQwen3.8 Max (0902)100%2026
Zhipu AI (Z.ai)GLM-4.783.3%2025
Zhipu AI (Z.ai)GLM-5.193.3%2026
Zhipu AI (Z.ai)GLM-5.3-Flash93.9%2026
NVIDIANemotron 3 Ultra86.7%2026
MiniMaxMiniMax-M326.7%2026
MiniMaxMiniMax-M371.1%2026
Ai2 (Allen Institute)Tulu 3 (Tülu 3) 70B4.4%2024
MicrosoftPhi-413.8%2024

Record per lab: FrontierMath (tiers 1 to 3) ?

0%25%50%75%100%20252026Anthropic: Claude Opus 4.1, 12.6%, 2025Anthropic: Claude Sonnet 4.5, 23.9%, 2025Anthropic: Claude Opus 4.5, 34.4%, 2025Anthropic: Claude Opus 4.6, 66%, 2026Anthropic: Claude Opus 4.7, 70.2%, 2026Anthropic: Claude Opus 4.8, 80%, 2026Anthropic: Claude Fable 5, 87%, 2026Anthropic: Claude Fable 5.1, 90.2%, 2026Anthropic: Claude Opus 5.5, 91.2%, 2026Anthropic: Claude Fable 5.2, 95.1%, 2026OpenAI: GPT-3.5 Turbo, 0%, 2024OpenAI: GPT-4 Turbo (Apr 2024), 0.7%, 2024OpenAI: o1, 8.4%, 2024OpenAI: o1, 10.2%, 2024OpenAI: o1, 14.7%, 2024OpenAI: o3-mini, 18.6%, 2025OpenAI: o3, 19.3%, 2025OpenAI: o4-mini, 28.8%, 2025OpenAI: o3, 29.8%, 2025OpenAI: o3, 33.3%, 2025OpenAI: o4-mini, 36.1%, 2025OpenAI: GPT-5, 37.2%, 2025OpenAI: GPT-5 mini, 46.7%, 2025OpenAI: GPT-5, 55.4%, 2025OpenAI: GPT-5 Pro, 55.8%, 2025OpenAI: GPT-5.2, 67.4%, 2025OpenAI: GPT-5.2 Pro, 74%, 2025OpenAI: GPT-5.4, 78.6%, 2026OpenAI: GPT-5.4 Pro, 82.5%, 2026OpenAI: GPT-5.5, 85.3%, 2026OpenAI: GPT-5.5 Pro, 87.7%, 2026OpenAI: GPT-5.6 Sol, 89.1%, 2026OpenAI: GPT-6 Astra, 93.7%, 2026Google: Gemini 2.5 Pro (Jun 2025), 24.6%, 2025Google: Gemini 3 Flash, 51.2%, 2025Google: Gemini 3.1 Pro, 59.6%, 2026Google: Gemini 3.5 Flash, 62.8%, 2026Google: Gemini 3.7 Flash, 71.6%, 2026Google: Gemini 4 Argon, 86%, 2026Meta: Muse Spark 1.3, 74%, 2026Meta: Muse Spark 1.3, 74.4%, 2026Meta: Muse Wonder, 82.1%, 2026Moonshot AI: Kimi K2.6, 57.2%, 2026Moonshot AI: Kimi K3, 72.2%, 2026Moonshot AI: Kimi K3.1 Pro, 78.2%, 2026Alibaba: Qwen3-Max, 18.9%, 2025Alibaba: Qwen3.5 397B-A17B, 29.5%, 2026Alibaba: Qwen3.5 397B-A17B, 31.2%, 2026Alibaba: Qwen 3.6 Plus, 32.3%, 2026Alibaba: Qwen 3.6 Plus, 38.2%, 2026Alibaba: Qwen3.7-Max, 64.6%, 2026Alibaba: Qwen 3.8 Max, 74.7%, 2026DeepSeek: DeepSeek-V4-Pro, 45.3%, 2026DeepSeek: DeepSeek V4 Flash 0731, 57.5%, 2026DeepSeek: DeepSeek V4 Pro 0813, 64.6%, 2026DeepSeek: DeepSeek V4.1 Flash, 67.4%, 2026DeepSeek: DeepSeek V4.1 Flash 1017, 71.2%, 2026Zhipu AI (Z.ai): GLM-5.1, 24.9%, 2026Zhipu AI (Z.ai): GLM-5.1, 36.8%, 2026Zhipu AI (Z.ai): GLM-5.2, 42.5%, 2026Zhipu AI (Z.ai): GLM-5.2, 54.7%, 2026Zhipu AI (Z.ai): GLM-5.2, 59.2%, 2026Zhipu AI (Z.ai): GLM-5.3, 68.8%, 2026xAI: Grok 4.20, 44.9%, 2026xAI: Grok 4.5, 57.2%, 2026xAI: Grok 4.6, 66%, 2026Mistral AI: Mistral Large 4, 58.6%, 2026
  • Anthropic 95.1%
  • OpenAI 93.7%
  • Google 86%
  • Meta 82.1%
  • Moonshot AI 78.2%
  • Alibaba 74.7%
  • DeepSeek 71.2%
  • Zhipu AI (Z.ai) 68.8%
  • xAI 66%
  • Mistral AI 58.6%

Source: Epoch AI, CC BY 4.0, latest version we checked. The 90% to 100% band marks the saturation zone.

Chart data table
LabModel%Dates in sources
AnthropicClaude Opus 4.112.6%2025
AnthropicClaude Sonnet 4.523.9%2025
AnthropicClaude Opus 4.534.4%2025
AnthropicClaude Opus 4.666%2026
AnthropicClaude Opus 4.770.2%2026
AnthropicClaude Opus 4.880%2026
AnthropicClaude Fable 587%2026
AnthropicClaude Fable 5.190.2%2026
AnthropicClaude Opus 5.591.2%2026
AnthropicClaude Fable 5.295.1%2026
Moonshot AIKimi K2.657.2%2026
Moonshot AIKimi K372.2%2026
Moonshot AIKimi K3.1 Pro78.2%2026
OpenAIGPT-3.5 Turbo0%2024
OpenAIGPT-4 Turbo (Apr 2024)0.7%2024
OpenAIo18.4%2024
OpenAIo110.2%2024
OpenAIo114.7%2024
OpenAIo3-mini18.6%2025
OpenAIo319.3%2025
OpenAIo4-mini28.8%2025
OpenAIo329.8%2025
OpenAIo333.3%2025
OpenAIo4-mini36.1%2025
OpenAIGPT-537.2%2025
OpenAIGPT-5 mini46.7%2025
OpenAIGPT-555.4%2025
OpenAIGPT-5 Pro55.8%2025
OpenAIGPT-5.267.4%2025
OpenAIGPT-5.2 Pro74%2025
OpenAIGPT-5.478.6%2026
OpenAIGPT-5.4 Pro82.5%2026
OpenAIGPT-5.585.3%2026
OpenAIGPT-5.5 Pro87.7%2026
OpenAIGPT-5.6 Sol89.1%2026
OpenAIGPT-6 Astra93.7%2026
DeepSeekDeepSeek-V4-Pro45.3%2026
DeepSeekDeepSeek V4 Flash 073157.5%2026
DeepSeekDeepSeek V4 Pro 081364.6%2026
DeepSeekDeepSeek V4.1 Flash67.4%2026
DeepSeekDeepSeek V4.1 Flash 101771.2%2026
MetaMuse Spark 1.374%2026
MetaMuse Spark 1.374.4%2026
MetaMuse Wonder82.1%2026
GoogleGemini 2.5 Pro (Jun 2025)24.6%2025
GoogleGemini 3 Flash51.2%2025
GoogleGemini 3.1 Pro59.6%2026
GoogleGemini 3.5 Flash62.8%2026
GoogleGemini 3.7 Flash71.6%2026
GoogleGemini 4 Argon86%2026
Mistral AIMistral Large 458.6%2026
xAIGrok 4.2044.9%2026
xAIGrok 4.557.2%2026
xAIGrok 4.666%2026
AlibabaQwen3-Max18.9%2025
AlibabaQwen3.5 397B-A17B29.5%2026
AlibabaQwen3.5 397B-A17B31.2%2026
AlibabaQwen 3.6 Plus32.3%2026
AlibabaQwen 3.6 Plus38.2%2026
AlibabaQwen3.7-Max64.6%2026
AlibabaQwen 3.8 Max74.7%2026
Zhipu AI (Z.ai)GLM-5.124.9%2026
Zhipu AI (Z.ai)GLM-5.136.8%2026
Zhipu AI (Z.ai)GLM-5.242.5%2026
Zhipu AI (Z.ai)GLM-5.254.7%2026
Zhipu AI (Z.ai)GLM-5.259.2%2026
Zhipu AI (Z.ai)GLM-5.368.8%2026

Record per lab: SimpleQA Verified ?

0%25%50%75%100%20252026Anthropic: Claude 3 Opus, 12.6%, 2024Anthropic: Claude Sonnet 4.5, 23.7%, 2025Anthropic: Claude Sonnet 4.5, 30.7%, 2025Anthropic: Claude Opus 4.5, 45.7%, 2025Anthropic: Claude Opus 4.6, 47%, 2026Anthropic: Claude Opus 4.7, 51.7%, 2026Anthropic: Claude Opus 4.8, 53%, 2026Anthropic: Claude Fable 5, 70.7%, 2026Anthropic: Claude Fable 5.1, 70.8%, 2026Anthropic: Claude Opus 5.5, 72.2%, 2026Anthropic: Claude Fable 5.2, 94%, 2026Google: Gemini 3 Flash, 66.8%, 2025Google: Gemini 3.1 Pro, 73.5%, 2026Google: Gemini 4 Argon, 76%, 2026OpenAI: GPT-4o mini, 8.3%, 2024OpenAI: GPT-4o, 26%, 2024OpenAI: o1, 41.1%, 2024OpenAI: o3, 49.4%, 2025OpenAI: GPT-5, 50.1%, 2025OpenAI: GPT-5.5, 63%, 2026OpenAI: GPT-5.6 Sol, 69.7%, 2026OpenAI: GPT-6 Astra, 75.6%, 2026Moonshot AI: Kimi K2.5, 34.3%, 2026Moonshot AI: Kimi K2.6, 34.9%, 2026Moonshot AI: Kimi K2.7 Code, 36.5%, 2026Moonshot AI: Kimi K3, 50.6%, 2026Moonshot AI: Kimi K3.1 Pro, 62.2%, 2026Meta: Muse Spark 1.1, 57.8%, 2026Meta: Muse Spark 1.2, 60.3%, 2026xAI: Grok 4.20, 30.2%, 2026xAI: Grok 4.3 Beta, 33.2%, 2026xAI: Grok 4.5, 48.3%, 2026xAI: Grok 4.6, 48.9%, 2026xAI: Grok 4.6, 49.3%, 2026xAI: Grok 4.7, 56%, 2026Alibaba: Qwen3-235B-A22B (Jul 2025), 40.4%, 2025Alibaba: Qwen3-Max, 48.7%, 2025Alibaba: Qwen 3.6 Max (Preview), 52%, 2026Alibaba: Qwen3.7-Max, 55.8%, 2026DeepSeek: DeepSeek-V4-Pro, 47%, 2026DeepSeek: DeepSeek V4 Pro 0813, 52.9%, 2026ByteDance: Doubao-Seed-1.6, 17.4%, 2025ByteDance: Doubao-Seed-2.0 Pro, 39.4%, 2026ByteDance: Doubao Seed 2.1 Pro, 44.7%, 2026Zhipu AI (Z.ai): GLM-4.7, 32.2%, 2025Zhipu AI (Z.ai): GLM-5.1, 34%, 2026Zhipu AI (Z.ai): GLM-5.2, 34.2%, 2026Zhipu AI (Z.ai): GLM-5.3, 41%, 2026Mistral AI: Mistral Large 4, 20%, 2026Mistral AI: Mistral Large 4, 22.4%, 2026
  • Anthropic 94%
  • Google 76%
  • OpenAI 75.6%
  • Moonshot AI 62.2%
  • Meta 60.3%
  • xAI 56%
  • Alibaba 55.8%
  • DeepSeek 52.9%
  • ByteDance 44.7%
  • Zhipu AI (Z.ai) 41%
  • Mistral AI 22.4%

Source: Epoch AI, CC BY 4.0, latest version we checked. The 90% to 100% band marks the saturation zone.

Chart data table
LabModel%Dates in sources
AnthropicClaude 3 Opus12.6%2024
AnthropicClaude Sonnet 4.523.7%2025
AnthropicClaude Sonnet 4.530.7%2025
AnthropicClaude Opus 4.545.7%2025
AnthropicClaude Opus 4.647%2026
AnthropicClaude Opus 4.751.7%2026
AnthropicClaude Opus 4.853%2026
AnthropicClaude Fable 570.7%2026
AnthropicClaude Fable 5.170.8%2026
AnthropicClaude Opus 5.572.2%2026
AnthropicClaude Fable 5.294%2026
Moonshot AIKimi K2.534.3%2026
Moonshot AIKimi K2.634.9%2026
Moonshot AIKimi K2.7 Code36.5%2026
Moonshot AIKimi K350.6%2026
Moonshot AIKimi K3.1 Pro62.2%2026
ByteDanceDoubao-Seed-1.617.4%2025
ByteDanceDoubao-Seed-2.0 Pro39.4%2026
ByteDanceDoubao Seed 2.1 Pro44.7%2026
Mistral AIMistral Large 420%2026
Mistral AIMistral Large 422.4%2026
OpenAIGPT-4o mini8.3%2024
OpenAIGPT-4o26%2024
OpenAIo141.1%2024
OpenAIo349.4%2025
OpenAIGPT-550.1%2025
OpenAIGPT-5.563%2026
OpenAIGPT-5.6 Sol69.7%2026
OpenAIGPT-6 Astra75.6%2026
DeepSeekDeepSeek-V4-Pro47%2026
DeepSeekDeepSeek V4 Pro 081352.9%2026
MetaMuse Spark 1.157.8%2026
MetaMuse Spark 1.260.3%2026
GoogleGemini 3 Flash66.8%2025
GoogleGemini 3.1 Pro73.5%2026
GoogleGemini 4 Argon76%2026
xAIGrok 4.2030.2%2026
xAIGrok 4.3 Beta33.2%2026
xAIGrok 4.548.3%2026
xAIGrok 4.648.9%2026
xAIGrok 4.649.3%2026
xAIGrok 4.756%2026
AlibabaQwen3-235B-A22B (Jul 2025)40.4%2025
AlibabaQwen3-Max48.7%2025
AlibabaQwen 3.6 Max (Preview)52%2026
AlibabaQwen3.7-Max55.8%2026
Zhipu AI (Z.ai)GLM-4.732.2%2025
Zhipu AI (Z.ai)GLM-5.134%2026
Zhipu AI (Z.ai)GLM-5.234.2%2026
Zhipu AI (Z.ai)GLM-5.341%2026

Record per lab: SWE-bench Verified ?

0%25%50%75%100%20252026Anthropic: Claude 3.7 Sonnet, 61%, 2025Anthropic: Claude Opus 4, 70.7%, 2025Anthropic: Claude Opus 4.1, 73.3%, 2025Anthropic: Claude Opus 4.5, 76.7%, 2025Anthropic: Claude Opus 4.6, 78.7%, 2026Anthropic: Claude Opus 4.7, 83.5%, 2026OpenAI: GPT-4o, 31%, 2024OpenAI: GPT-4.1, 48.5%, 2025OpenAI: o3, 62.3%, 2025OpenAI: GPT-5 mini, 64.7%, 2025OpenAI: GPT-5, 71.5%, 2025OpenAI: GPT-5, 73.6%, 2025OpenAI: GPT-5.2, 73.8%, 2025OpenAI: GPT-5.3 Codex, 74.8%, 2026OpenAI: GPT-5.4, 76.9%, 2026OpenAI: GPT-5.5, 80.6%, 2026Google: Gemini 2.5 Pro (Jun 2025), 57.6%, 2025Google: Gemini 3 Pro, 72.9%, 2025Google: Gemini 3 Flash, 75.4%, 2025Google: Gemini 3.1 Pro, 75.6%, 2026Google: Gemini 3.5 Flash, 79.3%, 2026Zhipu AI (Z.ai): GLM-5, 72.1%, 2026Zhipu AI (Z.ai): GLM-5.1, 74.2%, 2026Zhipu AI (Z.ai): GLM-5.2, 78.7%, 2026DeepSeek: DeepSeek-V4-Pro, 77.6%, 2026Alibaba: Qwen 3.6 Plus, 57.9%, 2026Alibaba: Qwen 3.6 Max (Preview), 76.7%, 2026Alibaba: Qwen3.7-Max, 77.3%, 2026Moonshot AI: Kimi K2.5, 73.8%, 2026Moonshot AI: Kimi K2.6, 76.7%, 2026
  • Anthropic 83.5%
  • OpenAI 80.6%
  • Google 79.3%
  • Zhipu AI (Z.ai) 78.7%
  • DeepSeek 77.6%
  • Alibaba 77.3%
  • Moonshot AI 76.7%

Source: Epoch AI, CC BY 4.0, latest version we checked. The 90% to 100% band marks the saturation zone.

Chart data table
LabModel%Dates in sources
Zhipu AI (Z.ai)GLM-572.1%2026
Zhipu AI (Z.ai)GLM-5.174.2%2026
Zhipu AI (Z.ai)GLM-5.278.7%2026
AlibabaQwen 3.6 Plus57.9%2026
AlibabaQwen 3.6 Max (Preview)76.7%2026
AlibabaQwen3.7-Max77.3%2026
DeepSeekDeepSeek-V4-Pro77.6%2026
GoogleGemini 2.5 Pro (Jun 2025)57.6%2025
GoogleGemini 3 Pro72.9%2025
GoogleGemini 3 Flash75.4%2025
GoogleGemini 3.1 Pro75.6%2026
GoogleGemini 3.5 Flash79.3%2026
Moonshot AIKimi K2.573.8%2026
Moonshot AIKimi K2.676.7%2026
OpenAIGPT-4o31%2024
OpenAIGPT-4.148.5%2025
OpenAIo362.3%2025
OpenAIGPT-5 mini64.7%2025
OpenAIGPT-571.5%2025
OpenAIGPT-573.6%2025
OpenAIGPT-5.273.8%2025
OpenAIGPT-5.3 Codex74.8%2026
OpenAIGPT-5.476.9%2026
OpenAIGPT-5.580.6%2026
AnthropicClaude 3.7 Sonnet61%2025
AnthropicClaude Opus 470.7%2025
AnthropicClaude Opus 4.173.3%2025
AnthropicClaude Opus 4.576.7%2025
AnthropicClaude Opus 4.678.7%2026
AnthropicClaude Opus 4.783.5%2026

Input price per 1M tokens by OpenRouter listing date (log scale) ?

$0.001$0.01$0.10$1.00$10.00$100.0020252026Alibaba: Qwen2.5 72B Instruct, US$ 0.36, 2024Alibaba: Qwen: Qwen2.5 7B Instruct, US$ 0.10, 2024Alibaba: Qwen2.5 Coder 32B Instruct, US$ 0.66, 2024Alibaba: Qwen: Qwen2.5 VL 72B Instruct, US$ 0.80, 2025Alibaba: Qwen: Qwen-Plus, US$ 0.26, 2025Alibaba: Qwen: Qwen3 30B A3B, US$ 0.12, 2025Alibaba: Qwen: Qwen3 14B, US$ 0.10, 2025Alibaba: Qwen: Qwen3 32B, US$ 0.08, 2025Alibaba: Qwen: Qwen3 235B A22B Instruct 2507, US$ 0.14, 2025Alibaba: Qwen: Qwen3 Coder 480B A35B, US$ 0.30, 2025Alibaba: Qwen: Qwen3 235B A22B Thinking 2507, US$ 0.45, 2025Alibaba: Qwen: Qwen3 30B A3B Instruct 2507, US$ 0.10, 2025Alibaba: Qwen: Qwen3 Coder 30B A3B Instruct, US$ 0.15, 2025Alibaba: Qwen: Qwen3 Next 80B A3B Instruct, US$ 0.09, 2025Alibaba: Qwen: Qwen3 Coder Flash, US$ 0.20, 2025Alibaba: Qwen: Qwen3 VL 235B A22B Instruct, US$ 0.21, 2025Alibaba: Qwen: Qwen3 VL 30B A3B Thinking, US$ 0.29, 2025Alibaba: Qwen: Qwen3 VL 30B A3B Instruct, US$ 0.15, 2025Alibaba: Qwen: Qwen3 VL 8B Instruct, US$ 0.25, 2025Alibaba: Qwen: Qwen3 Coder Next, US$ 0.12, 2026Alibaba: Qwen: Qwen3.5 Plus 2026-02-15, US$ 0.26, 2026Alibaba: Qwen: Qwen3.5 397B A17B, US$ 0.60, 2026Alibaba: Qwen: Qwen3.5-35B-A3B, US$ 0.14, 2026Alibaba: Qwen: Qwen3.5-27B, US$ 0.26, 2026Alibaba: Qwen: Qwen3.5-122B-A10B, US$ 0.26, 2026Alibaba: Qwen: Qwen3.5-Flash, US$ 0.065, 2026Alibaba: Qwen: Qwen3.5-9B, US$ 0.17, 2026Alibaba: Qwen: Qwen3.6 Plus, US$ 0.33, 2026Alibaba: Qwen: Qwen3.5 Plus 2026-04-20, US$ 0.30, 2026Alibaba: Qwen: Qwen3.6 Flash, US$ 0.19, 2026Alibaba: Qwen: Qwen3.6 35B A3B, US$ 0.10, 2026Alibaba: Qwen: Qwen3.6 27B, US$ 0.32, 2026Alibaba: Qwen: Qwen3.7 Max, US$ 1.48, 2026Alibaba: Qwen: Qwen3.7 Plus, US$ 0.32, 2026Alibaba: Qwen: Qwen3.7 Flash, US$ 0.03, 2026Alibaba: Qwen: Qwen3.8 2.4T A95B, US$ 2.00, 2026Alibaba: Qwen: Qwen3.8 27B, US$ 0.15, 2026Alibaba: Qwen: Qwen3.8 Flash, US$ 0.15, 2026Alibaba: Qwen: Qwen3.8 Max (0902), US$ 2.00, 2026Alibaba: Qwen: Qwen3.8 Omni Flash, US$ 0.15, 2026Alibaba: Qwen: Qwen3.8 Max Prime, US$ 4.00, 2026Anthropic: Anthropic: Claude Opus 4.1, US$ 15.00, 2025Anthropic: Anthropic: Claude Sonnet 4.5, US$ 3.00, 2025Anthropic: Anthropic: Claude Haiku 4.5, US$ 1.00, 2025Anthropic: Anthropic: Claude Opus 4.5, US$ 5.00, 2025Anthropic: Anthropic: Claude Opus 4.6, US$ 5.00, 2026Anthropic: Anthropic: Claude Sonnet 4.6, US$ 3.00, 2026Anthropic: Anthropic: Claude Opus 4.7, US$ 5.00, 2026Anthropic: Anthropic: Claude Opus 4.8, US$ 5.00, 2026Anthropic: Anthropic: Claude Fable 5, US$ 10.00, 2026Anthropic: Anthropic: Claude Sonnet 5, US$ 2.00, 2026Anthropic: Anthropic: Claude Opus 5, US$ 5.00, 2026Anthropic: Anthropic: Claude Fable 5.1, US$ 10.00, 2026Anthropic: Anthropic: Claude Opus 5.5, US$ 4.00, 2026Anthropic: Anthropic: Claude Sonnet 5.5, US$ 2.00, 2026Anthropic: Anthropic: Claude Fable 5.2, US$ 8.00, 2026DeepSeek: DeepSeek: DeepSeek V3, US$ 0.32, 2024DeepSeek: DeepSeek: DeepSeek V3 0324, US$ 0.29, 2025DeepSeek: DeepSeek: R1 0528, US$ 0.50, 2025DeepSeek: DeepSeek: DeepSeek V3.1, US$ 0.25, 2025DeepSeek: DeepSeek: DeepSeek V3.1 Terminus, US$ 0.27, 2025DeepSeek: DeepSeek: DeepSeek V3.2 Exp, US$ 0.27, 2025DeepSeek: DeepSeek: DeepSeek V3.2, US$ 0.26, 2025DeepSeek: DeepSeek: DeepSeek V4 Pro 0423, US$ 0.96, 2026DeepSeek: DeepSeek: DeepSeek V4 Flash 0423, US$ 0.091, 2026DeepSeek: DeepSeek: DeepSeek V4 Flash 0731, US$ 0.043, 2026DeepSeek: DeepSeek: DeepSeek V4 Pro 0813, US$ 0.11, 2026DeepSeek: DeepSeek: DeepSeek V4 Flash Vision Exp, US$ 0.44, 2026DeepSeek: DeepSeek: DeepSeek V4.1 Flash, US$ 0.30, 2026DeepSeek: DeepSeek: DeepSeek V4.1 Flash 1017, US$ 0.006, 2026Google: Google: Gemma 2 27B, US$ 0.65, 2024Google: Google: Gemma 3 27B, US$ 0.08, 2025Google: Google: Gemma 3 4B, US$ 0.05, 2025Google: Google: Gemma 3 12B, US$ 0.05, 2025Google: Google: Gemini 2.5 Pro Preview 06-05, US$ 1.25, 2025Google: Google: Gemini 2.5 Flash, US$ 0.30, 2025Google: Google: Gemini 2.5 Pro, US$ 1.25, 2025Google: Google: Gemini 2.5 Flash Lite, US$ 0.10, 2025Google: Google: Nano Banana (Gemini 2.5 Flash Image), US$ 0.30, 2025Google: Google: Nano Banana Pro (Gemini 3 Pro Image Preview), US$ 2.00, 2025Google: Google: Gemini 3 Flash Preview, US$ 0.50, 2025Google: Google: Gemini 3.1 Pro Preview, US$ 2.00, 2026Google: Google: Gemini 3.1 Pro Preview Custom Tools, US$ 2.00, 2026Google: Google: Nano Banana 2 (Gemini 3.1 Flash Image Preview), US$ 0.50, 2026Google: Google: Gemini 3.1 Flash Lite Preview, US$ 0.25, 2026Google: Google: Gemma 4 31B, US$ 0.20, 2026Google: Google: Gemma 4 26B A4B , US$ 0.08, 2026Google: Google: Gemini 3.1 Flash Lite, US$ 0.25, 2026Google: Google: Gemini 3.5 Flash, US$ 1.50, 2026Google: Google: Nano Banana 2 (Gemini 3.1 Flash Image), US$ 0.50, 2026Google: Google: Nano Banana Pro (Gemini 3 Pro Image), US$ 2.00, 2026Google: Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image), US$ 0.25, 2026Google: Google: Gemini 3.6 Flash, US$ 0.75, 2026Google: Google: Gemini 3.5 Flash Lite, US$ 0.30, 2026Google: Google: Gemini 3.7 Flash, US$ 0.75, 2026Google: Google: Gemini 3.8 Flash, US$ 0.82, 2026Google: Google: Nano Banana 2.1, US$ 1.50, 2026Google: Google: Gemini 4 Argon, US$ 2.00, 2026Google: Google: Gemini Omni 1.1 Flash, US$ 1.50, 2026Google: Google: Gemini Omni 1.2 Flash, US$ 1.50, 2026Meta: Meta: Llama 3.1 70B Instruct, US$ 0.72, 2024Meta: Meta: Llama 3.1 8B Instruct, US$ 0.02, 2024Meta: Meta: Llama 3.2 1B Instruct, US$ 0.027, 2024Meta: Meta: Llama 3.2 3B Instruct, US$ 0.05, 2024Meta: Meta: Llama 3.3 70B Instruct, US$ 0.71, 2024Meta: Meta: Llama 4 Maverick, US$ 0.35, 2025Meta: Meta: Llama 4 Scout, US$ 0.10, 2025Meta: Meta: Llama Guard 4 12B, US$ 0.18, 2025Meta: Meta: Muse Spark 1.1, US$ 1.25, 2026Meta: Meta: Muse Spark 1.2, US$ 1.25, 2026Meta: Meta: Muse Glimmer 30B, US$ 0.30, 2026Meta: Meta: Muse Spark 1.2 Contributor, US$ 0.10, 2026Meta: Meta: Muse Spark 1.3 Contributor, US$ 0.10, 2026Meta: Meta: Muse Spark 1.3, US$ 1.25, 2026Meta: Meta: Muse Wonder, US$ 4.00, 2026Meta: Meta: Muse Spark Open, US$ 1.25, 2026Meta: Meta: Muse Ember, US$ 0.10, 2026Mistral AI: Mistral Large, US$ 2.00, 2024Mistral AI: Mistral: Mixtral 8x22B Instruct, US$ 2.00, 2024Mistral AI: Mistral: Mistral Nemo, US$ 0.029, 2024Mistral AI: Mistral Large 2407, US$ 2.00, 2024Mistral AI: Mistral: Mistral Small 3, US$ 0.05, 2025Mistral AI: Mistral: Saba, US$ 0.20, 2025Mistral AI: Mistral: Mistral Small 3.1 24B, US$ 0.35, 2025Mistral AI: Mistral: Mistral Medium 3, US$ 0.40, 2025Mistral AI: Mistral: Mistral Small 3.2 24B, US$ 0.10, 2025Mistral AI: Mistral: Codestral 2508, US$ 0.30, 2025Mistral AI: Mistral: Mistral Medium 3.1, US$ 0.40, 2025Mistral AI: Mistral: Voxtral Small 24B 2507, US$ 0.10, 2025Mistral AI: Mistral: Mistral Large 3 2512, US$ 0.50, 2025Mistral AI: Mistral: Ministral 3 14B 2512, US$ 0.20, 2025Mistral AI: Mistral: Ministral 3 8B 2512, US$ 0.15, 2025Mistral AI: Mistral: Ministral 3 3B 2512, US$ 0.10, 2025Mistral AI: Mistral: Devstral 2 2512, US$ 0.40, 2025Mistral AI: Mistral: Mistral Small 4, US$ 0.15, 2026Mistral AI: Mistral: Mistral Medium 3.5, US$ 1.50, 2026Mistral AI: Mistral: Mistral Large 4, US$ 0.70, 2026Moonshot AI: MoonshotAI: Kimi K2 0711, US$ 0.57, 2025Moonshot AI: MoonshotAI: Kimi K2 0905, US$ 0.60, 2025Moonshot AI: MoonshotAI: Kimi K2 Thinking, US$ 0.60, 2025Moonshot AI: MoonshotAI: Kimi K2.5, US$ 0.53, 2026Moonshot AI: MoonshotAI: Kimi K2.6, US$ 0.65, 2026Moonshot AI: MoonshotAI: Kimi K2.7 Code, US$ 0.60, 2026Moonshot AI: MoonshotAI: Kimi K3.1 Pro, US$ 2.00, 2026OpenAI: OpenAI: GPT-3.5 Turbo (older v0613), US$ 1.00, 2024OpenAI: OpenAI: GPT-4o, US$ 2.50, 2024OpenAI: OpenAI: GPT-4o (2024-05-13), US$ 5.00, 2024OpenAI: OpenAI: GPT-4o-mini, US$ 0.15, 2024OpenAI: OpenAI: GPT-4o-mini (2024-07-18), US$ 0.15, 2024OpenAI: OpenAI: GPT-4o (2024-08-06), US$ 2.50, 2024OpenAI: OpenAI: GPT-4o (2024-11-20), US$ 2.50, 2024OpenAI: OpenAI: GPT-4.1, US$ 2.00, 2025OpenAI: OpenAI: GPT-4.1 Mini, US$ 0.40, 2025OpenAI: OpenAI: GPT-4.1 Nano, US$ 0.10, 2025OpenAI: OpenAI: o3, US$ 2.00, 2025OpenAI: OpenAI: o3 Pro, US$ 20.00, 2025OpenAI: OpenAI: gpt-oss-120b, US$ 0.15, 2025OpenAI: OpenAI: gpt-oss-20b, US$ 0.03, 2025OpenAI: OpenAI: GPT-5, US$ 1.25, 2025OpenAI: OpenAI: GPT-5 Mini, US$ 0.25, 2025OpenAI: OpenAI: GPT-5 Nano, US$ 0.05, 2025OpenAI: OpenAI: GPT-5 Pro, US$ 15.00, 2025OpenAI: OpenAI: GPT-5 Image, US$ 10.00, 2025OpenAI: OpenAI: GPT-5 Image Mini, US$ 2.50, 2025OpenAI: OpenAI: gpt-oss-safeguard-20b, US$ 0.075, 2025OpenAI: OpenAI: GPT-5.1, US$ 1.25, 2025OpenAI: OpenAI: GPT-5.1-Codex, US$ 1.25, 2025OpenAI: OpenAI: GPT-5.1-Codex-Mini, US$ 0.25, 2025OpenAI: OpenAI: GPT-5.1-Codex-Max, US$ 1.25, 2025OpenAI: OpenAI: GPT-5.2 Chat, US$ 1.75, 2025OpenAI: OpenAI: GPT-5.2 Pro, US$ 21.00, 2025OpenAI: OpenAI: GPT-5.2, US$ 1.75, 2025OpenAI: OpenAI: GPT-5.2-Codex, US$ 1.75, 2026OpenAI: OpenAI: GPT Audio, US$ 2.50, 2026OpenAI: OpenAI: GPT Audio Mini, US$ 0.60, 2026OpenAI: OpenAI: GPT-5.3-Codex, US$ 1.75, 2026OpenAI: OpenAI: GPT-5.4 Pro, US$ 30.00, 2026OpenAI: OpenAI: GPT-5.4, US$ 2.50, 2026OpenAI: OpenAI: GPT-5.4 Nano, US$ 0.20, 2026OpenAI: OpenAI: GPT-5.4 Mini, US$ 0.75, 2026OpenAI: OpenAI: GPT-5.4 Image 2, US$ 8.00, 2026OpenAI: OpenAI: GPT-5.5 Pro, US$ 30.00, 2026OpenAI: OpenAI: GPT-5.5, US$ 5.00, 2026OpenAI: OpenAI: GPT Chat Latest, US$ 5.00, 2026OpenAI: OpenAI: GPT-5.6 Luna Pro, US$ 0.20, 2026OpenAI: OpenAI: GPT-5.6 Luna, US$ 0.20, 2026OpenAI: OpenAI: GPT-5.6 Terra Pro, US$ 2.00, 2026OpenAI: OpenAI: GPT-5.6 Terra, US$ 2.00, 2026OpenAI: OpenAI: GPT-5.6 Sol Pro, US$ 4.00, 2026OpenAI: OpenAI: GPT-5.6 Sol, US$ 4.00, 2026OpenAI: OpenAI: GPT-6 Astra, US$ 10.00, 2026OpenAI: OpenAI: GPT-6 Astra Pro, US$ 10.00, 2026OpenAI: OpenAI: GPT-6 Luna Pro, US$ 0.10, 2026OpenAI: OpenAI: GPT-6 Luna, US$ 0.10, 2026OpenAI: OpenAI: GPT-6 Sol Pro, US$ 2.00, 2026OpenAI: OpenAI: GPT-6 Sol, US$ 2.00, 2026OpenAI: OpenAI: GPT-6.1 Sol Pro, US$ 2.00, 2026OpenAI: OpenAI: GPT-6.1 Sol, US$ 2.00, 2026OpenAI: OpenAI: GPT-6.1 Terra Pro, US$ 1.00, 2026OpenAI: OpenAI: GPT-6.1 Terra, US$ 1.00, 2026Zhipu AI (Z.ai): Z.ai: GLM 4.5, US$ 0.60, 2025Zhipu AI (Z.ai): Z.ai: GLM 4.5 Air, US$ 0.20, 2025Zhipu AI (Z.ai): Z.ai: GLM 4.5V, US$ 0.60, 2025Zhipu AI (Z.ai): Z.ai: GLM 4.6, US$ 0.50, 2025Zhipu AI (Z.ai): Z.ai: GLM 4.6V, US$ 0.30, 2025Zhipu AI (Z.ai): Z.ai: GLM 4.7, US$ 0.60, 2025Zhipu AI (Z.ai): Z.ai: GLM 4.7 Flash, US$ 0.06, 2026Zhipu AI (Z.ai): Z.ai: GLM 5, US$ 1.00, 2026Zhipu AI (Z.ai): Z.ai: GLM 5 Turbo, US$ 1.20, 2026Zhipu AI (Z.ai): Z.ai: GLM 5V Turbo, US$ 1.20, 2026Zhipu AI (Z.ai): Z.ai: GLM 5.1, US$ 1.40, 2026Zhipu AI (Z.ai): Z.ai: GLM 5.3, US$ 0.07, 2026Zhipu AI (Z.ai): Z.ai: GLM 5.3 Flash, US$ 0.04, 2026Zhipu AI (Z.ai): Z.ai: GLM 5.3 FlashX, US$ 0.37, 2026Zhipu AI (Z.ai): Z.ai: GLM 5.3 Prime, US$ 2.80, 2026Zhipu AI (Z.ai): Z.ai: GLM 5.3 FlashX Legal, US$ 0.37, 2026xAI: SpaceXAI: Grok 4.20 Multi-Agent, US$ 1.25, 2026xAI: SpaceXAI: Grok 4.20, US$ 1.25, 2026xAI: SpaceXAI: Grok 4.3, US$ 1.25, 2026xAI: SpaceXAI: Grok Build 0.1, US$ 1.00, 2026xAI: SpaceXAI: Grok 4.5, US$ 2.00, 2026xAI: SpaceXAI: Grok 4.6, US$ 2.00, 2026xAI: SpaceXAI: Grok 4.7, US$ 2.00, 2026
  • Alibaba
  • Anthropic
  • DeepSeek
  • Google
  • Meta
  • Mistral AI
  • Moonshot AI
  • OpenAI
  • Zhipu AI (Z.ai)
  • xAI

Source: OpenRouter, latest version we checked. Dates come from the sources: OpenRouter listing or Hugging Face repo creation, which may differ from the official launch.

Sources

  • Epoch AI, CC BY 4.0, latest version we checked.
  • OpenRouter, dados públicos da API; exports de benchmarks e rankings em CC BY 4.0, latest version we checked.

Dates come from the sources: OpenRouter listing or Hugging Face repo creation, which may differ from the official launch. Family and version were grouped by rule from the name.