GPQA Diamond
The official leaderboard hasn’t scored a model released after GPT-6.2 Astra (November 2026). Labs report their own numbers for one newer model: Inkling 2 Mini at 86.6%.
198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Saturated (about 94-96%). Several 2026 cards use it only as a substrate for CoT-controllability or monitor-evasion tests.
- Made by
- NYU (Rein et al.)
- Released
- November 2023
- Kind
- Science questions
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Another benchmark
- Refresh
- One fixed set
- Run by
- Anyone
- Based on
- GPQA Main
- In views
- Biology and medicine, Chemistry and materials
- Strongest model tested
- GPT-6.2 Astra #1 on the capability index
- Last in a launch post
- November 2026
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
In the authors’ words
We present GPQA, a challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. We ensure that the questions are high-quality and extremely difficult: experts who have or are pursuing PhDs in the corresponding domains reach 65% accuracy (74% when discounting clear mistakes the experts identified in retrospect), while highly skilled non-expert validators only reach 34% accuracy, despite spending on average over 30 minutes with unrestricted access to the web (i.e., the questions are "Google-proof"). The questions are also difficult for state-of-the-art AI systems, with our strongest GPT-4 based baseline achieving 39% accuracy. If we are to use future AI systems to help us answer very hard questions, for example, when developing new scientific knowledge, we need to develop scalable oversight methods that enable humans to supervise their outputs, which may be difficult even if the supervisors are themselves skilled and knowledgeable. The difficulty of GPQA both for skilled non-experts and frontier AI systems should enable realistic scalable oversight experiments, which we hope can help devise ways for human experts to reliably get truthful information from AI systems that surpass human capabilities.
Gemini 4 Carbon leads with 97%
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
Gemini 4 Carbon leads with 97% across all independent runs
Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Dotted bars are reruns by Epoch AI or another board, and each row names its source.
GPT-6.2 Astra leads with 97%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
- 97.0
Gemini 4 Carbon
- 95.0 Grok 4.6
- 95.0 GPT-6 Astra
- 95.0
Gemini 4 Argon
- 93.9
Gemini 3.1 Pro
- 93.9 GPT-6.1 Sol
- 92.9
Gemini 3.8 Flash
- 92.9 Claude Opus 5
- 92.6 Grok 4.7
- 91.9 GPT-6 Sol
- 91.9 Claude Sonnet 5.5
- 90.8 Claude Opus 5.5
- 89.9
Gemini 3.6 Flash
- 89.4
DeepSeek V4.1 Flash 1017
- 84.8 Claude Haiku 5.5
- 83.2
Gemma 4 26B A4B
- 82.8
Gemma 4 31B IT
- 81.8 GPT-6 Luna
- 80.8 GPT-5.6 Luna
- 79.3 GPT-5.6 Terra
- 1
Gemini 4 Carbon 97%
- 2 Grok 4.6 95%
- 3 GPT-6 Astra 95%
- 4
Gemini 4 Argon 95%
- 5
Gemini 3.1 Pro 93.9%
- 6 GPT-6.1 Sol 93.9%
- 7
Gemini 3.8 Flash 92.9%
- 8 Claude Opus 5 92.9%
- 9 Grok 4.7 92.6%
- 10 GPT-6 Sol 91.9%
- 11 Claude Sonnet 5.5 91.9%
- 12 Claude Opus 5.5 90.8%
- 13
Gemini 3.6 Flash 89.9%
- 14
DeepSeek V4.1 Flash 1017 89.4%
- 15 Claude Haiku 5.5 84.8%
- 16
Gemma 4 26B A4B 83.2%
- 17
Gemma 4 31B IT 82.8%
- 18 GPT-6 Luna 81.8%
- 19 GPT-5.6 Luna 80.8%
- 20 GPT-5.6 Terra 79.3%
Source: Official GPQA Diamond leaderboard
- 97.0
Gemini 4 Carbon
- 96.8 GPT-6.2 Astra
- 96.8
Gemini 4 Hydrogen
- 96.4
Gemini 4 Argon
- 96.0 Claude Fable 5.2
- 95.8 GPT-6 Astra
- 95.6 Claude Sonnet 5.5
- 95.4
Gemini 3.8 Flash
- 95.4 GPT-6.1 Sol
- 95.0 Grok 4.6
- 94.8
Gemini 3.7 Flash
- 94.6 GPT-5.4 Pro
- 94.4
Gemini 3.1 Pro
- 94.3 GPT-6 Sol
- 94.3 GPT-6.1 Terra
- 94.2 Kimi K3.1 Pro
- 94.1
Gemini 3.6 Flash
- 94.1 Claude Opus 5.5
- 94.0 GPT-5.5
- 93.9 GPT-5.5 Pro
- 93.9 Claude Fable 5
- 93.9 Claude Opus 5
- 93.5 GPT-5.6 Sol
- 93.4 Grok 4.5
- 93.3 GPT-5.6 Terra
- 93.3 GPT-5.4
- 93.2
DeepSeek V4.1 Flash 1017
- 93.1 Kimi K3
- 93.1
Doubao Seed Evolving
- 92.8
Gemini 3.5 Flash
- 92.7 Grok 4.7
- 92.7
Qwen 3.8 Max
- 92.6
Gemini 3 Pro
- 92.3
Qwen3.8 Max (0902)
- 91.9 GLM-5.2
- 91.7
DeepSeek V4 Pro 0813
- 91.6 GPT-5.6 Luna
- 91.4 GPT-5.2
- 91.0 Claude Opus 4.8
- 91.0
DeepSeek V4 Flash 0731
- 90.9
DeepSeek-V4-Pro
- 90.9
Qwen3.7-Max
- 90.9
MiniMax-M3
- 90.9 GLM-5.3
- 90.8 Kimi K2.6
- 90.5 Claude Opus 4.6
- 90.5 Claude Sonnet 5
- 90.5 GPT-6 Luna
- 90.2 Claude Opus 4.7
- 90.2 GLM-5.3-Flash
- 90.2
Mistral Large 4
- 89.9 GLM-5.1
- 89.8
DeepSeek V4.1 Flash
- 89.8
Muse Spark
- 89.7 Claude Haiku 5.5
- 89.6
Doubao Seed 2.1 Turbo
- 89.4
Gemini 3 Flash
- 89.3 Grok 4.20
- 89.3
Doubao Seed 2.1 Pro
- 88.8 Grok 4.3 Beta
- 88.6
Muse Spark Open
- 88.5
Inkling-Small
- 88.4
Qwen 3.6 Plus
- 88.3
Inkling
- 88.2
Doubao-Seed-2.0 Pro
- 88.1
Doubao Seed 2.0 Lite
- 88.1
Mistral Large 4 Preview
- 87.9
Qwen3.7-Plus
- 87.9 Kimi K2.7 Code
- 87.8 GLM-5
- 87.6 GPT-5.1
- 87.6 Kimi K2.5
- 87.4 Claude Sonnet 4.6
- 87.4
Qwen 3.6 Max (Preview)
- 87.0 Grok 4
- 86.9 GPT-5.4 Mini
- 86.4
Qwen3.5 397B-A17B
- 86.2 GPT-5
- 86.0 Claude Opus 4.5
- 85.9
Qwen3.6 27B
- 85.3
Nemotron 3 Ultra
- 85.3
Gemini 2.5 Pro
- 84.8
Qwen 3.5 Plus (hosted 397B-A17B)
- 84.8
Qwen 3.6 35B-A3B
- 84.2 Kimi K2 Thinking
- 83.8
Gemini 2.5 Pro
- 83.5
Qwen3.5-35B-A3B
- 83.4
DeepSeek-V3.2
- 83.3 GLM-4.7
- 83.3
Qwen 3.6 Flash
- 83.3
Gemini 3.5 Flash-Lite
- 83.2
Gemma 4 26B A4B
- 82.8
Gemma 4 31B IT
- 82.5 GPT-5.5 Instant
- 82.3 Claude Sonnet 4.5
- 82.3
Qwen 3.5 Flash (hosted 35B-A3B)
- 82.3
Qwen3.7 Flash
- 81.8 o3
- 81.8
Gemini 3.1 Flash-Lite
- 80.0
Qwen3-235B-A22B-Thinking
- 79.7 Claude 3.7 Sonnet
- 79.6 o4-mini
- 79.2 Claude Sonnet 4
- 79.0
Qwen3.5-9B
- 78.5 GPT-5.4 Nano
- 77.6
Doubao-Seed-1.6
- 77.3 Claude Opus 4.1
- 77.0 o3-mini
- 76.8 o1
- 76.3
DeepSeek-R1
- 76.3 Grok-3 mini
- 76.3 Claude Opus 4
- 75.8 Grok 3
- 75.8 gpt-oss-120b
- 75.0 GPT-5 mini
- 72.6
Qwen3-Max
- 71.7
DeepSeek-R1
- 71.5
seed-oss-36b-instruct
- 71.2 Claude Haiku 4.5
- 70.7
Qwen3-235B-A22B
- 70.1
Qwen3-30B-A3B-Thinking
- 69.4 GPT-5 nano
- 68.7 GPT-4.5
- 67.6
DeepSeek-V3
- 67.0
Llama 4 Maverick
- 66.9 GPT-4.1
- 66.7
Gemini 2.5 Pro
- 65.8 GPT-4.1 mini
- 65.7
Qwen3-32B
- 65.7
Gemini 2.0 Pro
- 65.4
QWQ-Plus
- 65.3
QwQ-32B
- 64.1
DeepSeek-R1-Distill-Qwen-32B
- 64.1
Gemini 2.0 Flash
- 63.8
Qwen3-14B
- 62.4 o1-mini
- 61.7
Qwen3-30B-A3B
- 60.8 gpt-oss-20b
- 60.5 glm-4.7-flash
- 59.5
Mistral Medium 3
- 57.2
Gemini 1.5 Pro
- 57.1
Gemini 2.0 Flash Thinking
- 56.8
Qwen3-8B
- 56.5
DeepSeek-V3
- 56.1
Qwen2.5-Max
- 56.1
Phi-4
- 56.1
Magistral Small 1.0
- 55.7
DeepSeek-R1-Distill-Llama-70B
- 55.6
Qwen3-30B-A3B-Instruct
- 55.3 Claude 3.5 Sonnet (October 2024)
- 54.0 Claude 3.5 Sonnet
- 53.8 Grok-2
- 52.3
Qwen3-4B
- 51.8
Llama 4 Scout
- 51.3
Mistral Large 2
- 50.9
Llama 3.1-405B
- 50.3 o1-preview
- 49.2 GPT-4o
- 49.1
Qwen2.5-72B
- 49.0
Mistral Small 3.2
- 49.0
Mistral Large 2
- 48.9 GPT-4.1 nano
- 48.9 GPT-4o
- 48.1
Qwen Plus
- 47.9 GPT-4o
- 47.7
Gemma 3 27B
- 47.6
Magistral Small 1.2
- 47.5
Mistral Small 3.1
- 47.4
Llama 3.3 70B
- 47.3
Gemini 1.5 Flash
- 47.3
Mistral Small 3
- 47.2 Claude 3 Opus
- 46.6 GPT-4 Turbo
- 46.3
Tulu 3 (Tülu 3) 70B
- 46.1
Qwen2.5-32B
- 45.9
Gemini 1.5 Pro
- 45.8 qwen3-4b-instruct-2507
- 45.1 glm-4.7-flash_none
- 44.7
DeepSeek-R1-Distill-Qwen-14B
- 44.2
Llama 3.1-70B
- 43.4
WizardLM-2 8x22B
- 42.4 GPT-4 Turbo
- 41.8
Qwen-Turbo
- 41.0
Llama 3.2 90B
- 40.8
Qwen2-72B
- 40.6 Claude 3 Sonnet
- 40.6
Llama 3-70B
- 40.4
Gemini 1.5 Flash
- 39.5
Gemma 3 12B
- 38.8
Mistral Large
- 38.1 Claude 3.5 Haiku
- 38.0
Qwen3-1.7B
- 37.7 GPT-4o mini
- 37.5 Hermes 2 Theta Llama-3 70B
- 36.5
Gemma 2 27B
- 36.3 Claude 3 Haiku
- 35.7 GPT-4
- 35.5
Qwen2.5-7B
- 34.7 Claude 2
- 34.1
Mixtral 8x22B
- 34.0
Gemini 1.0 Pro
- 33.9 Eurus-2-7B-PRIME
- 33.6
DeepSeek-R1-Distill-Qwen-1.5B
- 33.0 Claude 2.1
- 33.0
Gemini 1.5 Flash 8B
- 32.9
DBRX
- 32.0
Yi-1.5-34B
- 30.7
Qwen1.5-32B
- 30.6 GPT-4
- 30.6
Mixtral 8x7B
- 29.9
Mistral NeMo
- 28.8
Qwen1.5-72B
- 28.3 granite-4.0-micro
- 28.0 GPT-3.5 Turbo
- 27.6
phi-3-medium 14B
- 27.5
Gemma 2 9B
- 27.2 GPT-3.5 Turbo
- 27.1
Ministral 8B
- 27.0
Llama 3.1-8B
- 26.3
Llama 2-70B
- 26.1
Llama 3-8B
- 25.3
Ministral 3B
- 24.6
DeepSeek LLM 67B
- 24.0 granite-4.0-1b
- 23.9
Llama 3.2 1B
- 23.2
Gemma 3 4B
- 19.9
Gemma 3 1B
- 15.2
Mistral 7B v0.3
- 14.7
Yi-34B
- 11.2 granite-4.0-350m
- 9.28 deepseek-r1-0528-qwen3-8b
- 1
Gemini 4 Carbon 97%
- 2 GPT-6.2 Astra max 96.8%
- 3
Gemini 4 Hydrogen xhigh 96.8%
- 4
Gemini 4 Argon high 96.4%
- 5 Claude Fable 5.2 with fallback, max 96%
- 6 GPT-6 Astra max 95.8%
- 7 Claude Sonnet 5.5 max 95.6%
- 8
Gemini 3.8 Flash high 95.4%
- 9 GPT-6.1 Sol max 95.4%
- 10 Grok 4.6 95%
- 11
Gemini 3.7 Flash high 94.8%
- 12 GPT-5.4 Pro xhigh 94.6%
- 13
Gemini 3.1 Pro high 94.4%
- 14 GPT-6 Sol max 94.3%
- 15 GPT-6.1 Terra max 94.3%
- 16 Kimi K3.1 Pro Together, max 94.2%
- 17
Gemini 3.6 Flash high 94.1%
- 18 Claude Opus 5.5 with fallback, max 94.1%
- 19 GPT-5.5 xhigh 94%
- 20 GPT-5.5 Pro xhigh 93.9%
- 21 Claude Fable 5 with fallback, max 93.9%
- 22 Claude Opus 5 max 93.9%
- 23 GPT-5.6 Sol max 93.5%
- 24 Grok 4.5 high 93.4%
- 25 GPT-5.6 Terra max 93.3%
- 26 GPT-5.4 xhigh 93.3%
- 27
DeepSeek V4.1 Flash 1017 max 93.2%
- 28 Kimi K3 max 93.1%
- 29
Doubao Seed Evolving Nov 2026, high 93.1%
- 30
Gemini 3.5 Flash high 92.8%
- 31 Grok 4.7 xhigh 92.7%
- 32
Qwen 3.8 Max xhigh 92.7%
- 33
Gemini 3 Pro 92.6%
- 34
Qwen3.8 Max (0902) xhigh 92.3%
- 35 GLM-5.2 max 91.9%
- 36
DeepSeek V4 Pro 0813 max 91.7%
- 37 GPT-5.6 Luna max 91.6%
- 38 GPT-5.2 xhigh 91.4%
- 39 Claude Opus 4.8 max 91%
- 40
DeepSeek V4 Flash 0731 max 91%
- 41
DeepSeek-V4-Pro high 90.9%
- 42
Qwen3.7-Max 90.9%
- 43
MiniMax-M3 90.9%
- 44 GLM-5.3 max 90.9%
- 45 Kimi K2.6 90.8%
- 46 Claude Opus 4.6 32k thinking 90.5%
- 47 Claude Sonnet 5 xhigh 90.5%
- 48 GPT-6 Luna max 90.5%
- 49 Claude Opus 4.7 xhigh 90.2%
- 50 GLM-5.3-Flash max 90.2%
- 51
Mistral Large 4 high 90.2%
- 52 GLM-5.1 89.9%
- 53
DeepSeek V4.1 Flash max 89.8%
- 54
Muse Spark 89.8%
- 55 Claude Haiku 5.5 max 89.7%
- 56
Doubao Seed 2.1 Turbo high 89.6%
- 57
Gemini 3 Flash high 89.4%
- 58 Grok 4.20 89.3%
- 59
Doubao Seed 2.1 Pro high 89.3%
- 60 Grok 4.3 Beta high 88.8%
- 61
Muse Spark Open xhigh 88.6%
- 62
Inkling-Small xhigh 88.5%
- 63
Qwen 3.6 Plus 2026-04-02 88.4%
- 64
Inkling xhigh 88.3%
- 65
Doubao-Seed-2.0 Pro high 88.2%
- 66
Doubao Seed 2.0 Lite Apr 2026, high 88.1%
- 67
Mistral Large 4 Preview high 88.1%
- 68
Qwen3.7-Plus 87.9%
- 69 Kimi K2.7 Code 87.9%
- 70 GLM-5 87.8%
- 71 GPT-5.1 high 87.6%
- 72 Kimi K2.5 Fireworks 87.6%
- 73 Claude Sonnet 4.6 32k thinking 87.4%
- 74
Qwen 3.6 Max (Preview) 87.4%
- 75 Grok 4 87%
- 76 GPT-5.4 Mini xhigh 86.9%
- 77
Qwen3.5 397B-A17B none 86.4%
- 78 GPT-5 high 86.2%
- 79 Claude Opus 4.5 32k thinking 86%
- 80
Qwen3.6 27B 85.9%
- 81
Nemotron 3 Ultra 85.3%
- 82
Gemini 2.5 Pro Jun 2025 85.3%
- 83
Qwen 3.5 Plus (hosted 397B-A17B) 84.8%
- 84
Qwen 3.6 35B-A3B none 84.8%
- 85 Kimi K2 Thinking 84.2%
- 86
Gemini 2.5 Pro Mar 2025 83.8%
- 87
Qwen3.5-35B-A3B 83.5%
- 88
DeepSeek-V3.2 Thinking 83.4%
- 89 GLM-4.7 83.3%
- 90
Qwen 3.6 Flash 2026-04-16 83.3%
- 91
Gemini 3.5 Flash-Lite high 83.3%
- 92
Gemma 4 26B A4B 83.2%
- 93
Gemma 4 31B IT 82.8%
- 94 GPT-5.5 Instant 82.5%
- 95 Claude Sonnet 4.5 59k thinking 82.3%
- 96
Qwen 3.5 Flash (hosted 35B-A3B) 82.3%
- 97
Qwen3.7 Flash 82.3%
- 98 o3 high 81.8%
- 99
Gemini 3.1 Flash-Lite high 81.8%
- 100
Qwen3-235B-A22B-Thinking Jul 2025 80%
- 101 Claude 3.7 Sonnet 64k thinking 79.7%
- 102 o4-mini high 79.6%
- 103 Claude Sonnet 4 59K 79.2%
- 104
Qwen3.5-9B 79%
- 105 GPT-5.4 Nano high 78.5%
- 106
Doubao-Seed-1.6 thinking 77.6%
- 107 Claude Opus 4.1 16K 77.3%
- 108 o3-mini high 77%
- 109 o1 high 76.8%
- 110
DeepSeek-R1 May 2025 76.3%
- 111 Grok-3 mini low 76.3%
- 112 Claude Opus 4 16K 76.3%
- 113 Grok 3 75.8%
- 114 gpt-oss-120b high 75.8%
- 115 GPT-5 mini high 75%
- 116
Qwen3-Max 72.6%
- 117
DeepSeek-R1 71.7%
- 118
seed-oss-36b-instruct 71.5%
- 119 Claude Haiku 4.5 32K 71.2%
- 120
Qwen3-235B-A22B 70.7%
- 121
Qwen3-30B-A3B-Thinking Jul 2025 70.1%
- 122 GPT-5 nano high 69.4%
- 123 GPT-4.5 68.7%
- 124
DeepSeek-V3 Mar 2025 67.6%
- 125
Llama 4 Maverick FP8 67%
- 126 GPT-4.1 66.9%
- 127
Gemini 2.5 Pro May 2025 66.7%
- 128 GPT-4.1 mini 65.8%
- 129
Qwen3-32B 65.7%
- 130
Gemini 2.0 Pro 65.7%
- 131
QWQ-Plus 65.4%
- 132
QwQ-32B 65.3%
- 133
DeepSeek-R1-Distill-Qwen-32B 64.1%
- 134
Gemini 2.0 Flash Feb 2025 64.1%
- 135
Qwen3-14B 63.8%
- 136 o1-mini high 62.4%
- 137
Qwen3-30B-A3B 61.7%
- 138 gpt-oss-20b medium 60.8%
- 139 glm-4.7-flash 60.5%
- 140
Mistral Medium 3 59.5%
- 141
Gemini 1.5 Pro Sep 2024 57.2%
- 142
Gemini 2.0 Flash Thinking Jan 2025 57.1%
- 143
Qwen3-8B 56.8%
- 144
DeepSeek-V3 56.5%
- 145
Qwen2.5-Max 56.1%
- 146
Phi-4 56.1%
- 147
Magistral Small 1.0 56.1%
- 148
DeepSeek-R1-Distill-Llama-70B 55.7%
- 149
Qwen3-30B-A3B-Instruct Jul 2025 55.6%
- 150 Claude 3.5 Sonnet (October 2024) 55.3%
- 151 Claude 3.5 Sonnet Jun 2024 54%
- 152 Grok-2 Dec 2024 53.8%
- 153
Qwen3-4B 52.3%
- 154
Llama 4 Scout 51.8%
- 155
Mistral Large 2 Nov 2024 51.3%
- 156
Llama 3.1-405B 50.9%
- 157 o1-preview 50.3%
- 158 GPT-4o Aug 2024 49.2%
- 159
Qwen2.5-72B 49.1%
- 160
Mistral Small 3.2 49%
- 161
Mistral Large 2 Jul 2024 49%
- 162 GPT-4.1 nano 48.9%
- 163 GPT-4o May 2024 48.9%
- 164
Qwen Plus Jan 2025 48.1%
- 165 GPT-4o Nov 2024 47.9%
- 166
Gemma 3 27B 47.7%
- 167
Magistral Small 1.2 47.6%
- 168
Mistral Small 3.1 47.5%
- 169
Llama 3.3 70B 47.4%
- 170
Gemini 1.5 Flash Sep 2024 47.3%
- 171
Mistral Small 3 47.3%
- 172 Claude 3 Opus 47.2%
- 173 GPT-4 Turbo Apr 2024 46.6%
- 174
Tulu 3 (Tülu 3) 70B 46.3%
- 175
Qwen2.5-32B 46.1%
- 176
Gemini 1.5 Pro May 2024 45.9%
- 177 qwen3-4b-instruct-2507 45.8%
- 178 glm-4.7-flash_none none 45.1%
- 179
DeepSeek-R1-Distill-Qwen-14B 44.7%
- 180
Llama 3.1-70B 44.2%
- 181
WizardLM-2 8x22B 43.4%
- 182 GPT-4 Turbo Nov 2023 42.4%
- 183
Qwen-Turbo 41.8%
- 184
Llama 3.2 90B 41%
- 185
Qwen2-72B 40.8%
- 186 Claude 3 Sonnet 40.6%
- 187
Llama 3-70B 40.6%
- 188
Gemini 1.5 Flash May 2024 40.4%
- 189
Gemma 3 12B 39.5%
- 190
Mistral Large 38.8%
- 191 Claude 3.5 Haiku Oct 2024 38.1%
- 192
Qwen3-1.7B 38%
- 193 GPT-4o mini 37.7%
- 194 Hermes 2 Theta Llama-3 70B 37.5%
- 195
Gemma 2 27B 36.5%
- 196 Claude 3 Haiku 36.3%
- 197 GPT-4 Mar 2023 35.7%
- 198
Qwen2.5-7B 35.5%
- 199 Claude 2 34.7%
- 200
Mixtral 8x22B 34.1%
- 201
Gemini 1.0 Pro 34%
- 202 Eurus-2-7B-PRIME 33.9%
- 203
DeepSeek-R1-Distill-Qwen-1.5B 33.6%
- 204 Claude 2.1 33%
- 205
Gemini 1.5 Flash 8B 33%
- 206
DBRX 32.9%
- 207
Yi-1.5-34B 32%
- 208
Qwen1.5-32B 30.7%
- 209 GPT-4 Jun 2023 30.6%
- 210
Mixtral 8x7B 30.6%
- 211
Mistral NeMo 29.9%
- 212
Qwen1.5-72B 28.8%
- 213 granite-4.0-micro 28.3%
- 214 GPT-3.5 Turbo Nov 2023 28%
- 215
phi-3-medium 14B 27.6%
- 216
Gemma 2 9B 27.5%
- 217 GPT-3.5 Turbo Jan 2024 27.2%
- 218
Ministral 8B 27.1%
- 219
Llama 3.1-8B 27%
- 220
Llama 2-70B 26.3%
- 221
Llama 3-8B 26.1%
- 222
Ministral 3B 25.3%
- 223
DeepSeek LLM 67B 24.6%
- 224 granite-4.0-1b 24%
- 225
Llama 3.2 1B 23.9%
- 226
Gemma 3 4B 23.2%
- 227
Gemma 3 1B 19.9%
- 228
Mistral 7B v0.3 15.2%
- 229
Yi-34B 14.7%
- 230 granite-4.0-350m 11.2%
- 231 deepseek-r1-0528-qwen3-8b 9.28%
Sources: Official GPQA Diamond leaderboard and Epoch AI runs (CC BY 4.0)
- 97.0 GPT-6.2 Astra
- 97.0
Gemini 4 Carbon
- 96.8
Gemini 4 Hydrogen
- 96.4
Gemini 4 Argon
- 96.0 Claude Fable 5.2
- 96.0 GPT-6 Astra
- 96.0 GPT-6.2 Sol
- 95.6 Claude Sonnet 5.5
- 95.4
Gemini 3.8 Flash
- 95.4 GPT-6.1 Sol
- 95.0 Grok 4.6
- 94.8
Gemini 3.7 Flash
- 94.6 GPT-5.4 Pro
- 94.6 GPT-5.6 Sol
- 94.5 Claude Mythos Preview
- 94.4
Gemini 3.1 Pro
- 94.3 GPT-6 Sol
- 94.3 GPT-6.1 Terra
- 94.2 Claude Opus 4.7
- 94.2 Kimi K3.1 Pro
- 94.1
Gemini 3.6 Flash
- 94.1 Claude Opus 5.5
- 94.0 GPT-5.5
- 93.9 GPT-5.5 Pro
- 93.9 Claude Fable 5
- 93.9 Claude Opus 5
- 93.6 Claude Opus 4.8
- 93.5 Kimi K3
- 93.5
Step 5 Preview
- 93.4 Grok 4.5
- 93.3 GPT-5.6 Terra
- 93.3 GPT-5.4
- 93.2 GPT-5.2 Pro
- 93.2
DeepSeek V4.1 Flash 1017
- 93.1
Doubao Seed Evolving
- 93.0
Inkling 2 Small
- 92.8
Gemini 3.5 Flash
- 92.7 Grok 4.7
- 92.7
Qwen 3.8 Max
- 92.6
Gemini 3 Pro
- 92.4 GPT-5.2
- 92.4
Qwen3.7-Max
- 92.3 GPT-5.6 Luna
- 92.3
Tencent Hy4 preview
- 92.3
Qwen3.8 Max (0902)
- 91.9 GLM-5.2
- 91.7
Qwen3.8 Flash
- 91.7
DeepSeek V4 Pro 0813
- 91.3 Claude Opus 4.6
- 91.0
DeepSeek V4 Flash 0731
- 91.0
ERNIE 5.1
- 91.0
Qwen3.8-Omni-Flash
- 90.9
DeepSeek-V4-Pro
- 90.9
MiniMax-M3
- 90.9 GLM-5.3
- 90.9
DeepSeek V4.1 Flash
- 90.8 Kimi K2.6
- 90.5 Claude Sonnet 5
- 90.5 Beam 1.0
- 90.5 GPT-6 Luna
- 90.4
Gemini 3 Flash
- 90.4
Qwen 3.6 Plus
- 90.4
Tencent Hy3
- 90.3
Qwen3.7-Plus
- 90.2 GLM-5.3-Flash
- 90.2
Mistral Large 4
- 89.9 Claude Sonnet 4.6
- 89.9 GLM-5.1
- 89.8
Muse Spark
- 89.8
MiniMax-M2.7
- 89.7 Claude Haiku 5.5
- 89.6
Doubao Seed 2.1 Turbo
- 89.5
Nemotron 3.5 Super VL
- 89.5
Inkling-Small
- 89.4 GPT-5
- 89.3 Grok 4.20
- 89.3
Doubao Seed 2.1 Pro
- 89.2 Mercury 3
- 89.2
Qwen 3.8 27B
- 89.0
Solar Pro 4
- 89.0
Muse Spark Open
- 88.9
Seed2.0 Pro
- 88.9
LongCat-2.0
- 88.8 Grok 4.3 Beta
- 88.4 Grok 4 Heavy
- 88.4
Qwen3.5 397B-A17B
- 88.4
Seed2.0 Lite (0428)
- 88.3
Ring-2.6-1T
- 88.3
Inkling
- 88.2
Doubao-Seed-2.0 Pro
- 88.1
Doubao Seed 2.0 Lite
- 88.1 GPT-5.1
- 88.1
DeepSeek-V4-Flash
- 88.1
Mistral Large 4 Preview
- 88.0 GPT-5.4 Mini
- 87.9 Kimi K2.7 Code
- 87.8 GLM-5
- 87.8
Qwen3.6 27B
- 87.6 Kimi K2.5
- 87.5 Grok 4
- 87.4
Qwen3-Max-2026-01-23
- 87.4
Qwen 3.6 Max (Preview)
- 87.2
Tencent Hy3 preview
- 87.0 Claude Opus 4.5
- 87.0
Nemotron 3 Ultra
- 86.9
Gemini 3.1 Flash-Lite
- 86.6
Qwen3.5-122B-A10B
- 86.6
Inkling 2 Mini
- 86.4
Nemotron 3.5 Bolt 74B-A9B
- 86.4
Gemini 2.5 Pro
- 86.4
ERNIE 5.0
- 86.3
Solar Open 2
- 86.0
Qwen 3.6 35B-A3B
- 85.7 Grok 4 Fast
- 85.7
DeepSeek-V3.2-Speciale
- 85.7 GLM-4.7
- 85.6 GPT-5.5 Instant
- 85.5
Qwen3.5-27B
- 85.2
LongCat-Flash-Thinking-2601
- 85.2
MiniMax-M2.5
- 84.8
Qwen 3.5 Plus (hosted 397B-A17B)
- 84.8 Claude 3.7 Sonnet
- 84.6 Grok 3
- 84.6
MAI-Code-1-Flash
- 84.5 Kimi K2 Thinking
- 84.3 MiMo-V2-Flash
- 84.3
Gemma 4 31B IT
- 84.2
Qwen3.5-35B-A3B
- 84.2
MAI-Thinking-1
- 84.0 Grok-3 mini
- 84.0
Gemini 2.5 Pro
- 84.0 o3-pro
- 83.9
Qwen3.5-Omni-Plus
- 83.8 Claude Sonnet 4
- 83.8
Seed1.8
- 83.5
Step 3.5 Flash
- 83.5
Muse Glimmer
- 83.4
DeepSeek-V3.2
- 83.4 Claude Sonnet 4.5
- 83.3
Qwen 3.6 Flash
- 83.3
Gemini 3.5 Flash-Lite
- 83.3 o3
- 83.3 Claude Opus 4
- 83.3
gemini-robotics-er-1.5-preview
- 83.2
Gemma 4 26B A4B
- 83.0
MiniMax-M2.1
- 82.8
Gemini 2.5 Flash
- 82.8 GPT-5.4 Nano
- 82.3
Qwen 3.5 Flash (hosted 35B-A3B)
- 82.3
Qwen3.7 Flash
- 82.3 GPT-5 mini
- 82.2
K-EXAONE 2.0
- 81.7
Qwen3.5-9B
- 81.5
Seed1.6-Thinking
- 81.5
LongCat-Flash-Thinking
- 81.4 o4-mini
- 81.4
Nova 2.0 Pro Preview
- 81.1
Qwen3-235B-A22B-Thinking
- 81.0
DeepSeek-R1
- 81.0 GLM-4.6
- 80.9 Claude Opus 4.1
- 80.8
Gemini 2.5 Flash
- 80.7
DeepSeek-V3.1-Terminus
- 80.5
EXAONE 4.5 33B
- 80.1 gpt-oss-120b
- 80.1
DeepSeek-V3.1
- 79.9
DeepSeek-V3.2-Exp
- 79.7 o3-mini
- 79.6
Nova 2 Lite
- 79.2
Nemotron 3 Super
- 79.1 GLM-4.5
- 79.1
K-EXAONE
- 79.0 Mercury 2.5
- 78.8
Gemma 4 12B
- 78.7 Sarvam 105B
- 78.6
Ring-1T
- 78.3 o1-preview
- 78.3
Gemini 2.5 Flash
- 78.0
MiniMax-M2
- 77.7
Cogito v2.1
- 77.6
Doubao-Seed-1.6
- 77.3
Seed-Thinking-v1.5
- 77.3
Intern-S1
- 77.2
Qwen3-Next-80B-A3B
- 77.1
Qwen3-VL-235B-A22B
- 77.1
Spark-X2.6-4B
- 76.8 o1
- 76.5
Nemotron 3.5 Lightning 30B-A3B
- 76.3
Trinity Large Thinking
- 76.3
Magistral Medium 1.2
- 76.2
Qwen3.5-4B
- 76.2
Ling-2.6-1T
- 76.1
Nemotron-Cascade 2 30B-A3B
- 76.0
Cogito v2 Preview 671B MoE
- 76.0
Llama-3.1-Nemotron-Ultra-253B-v1
- 75.4
EXAONE 4.0 32B
- 75.2 glm-4.7-flash
- 75.1 Kimi K2
- 75.0 GLM-4.5-Air
- 74.5
Qwen3-Coder-Next
- 74.4 LFM2.5-24B-A2B
- 74.4
LongCat-Flash-Omni
- 74.4 INTELLECT-3
- 74.2
Gemini 2.0 Flash Thinking
- 74.0 Mercury 2
- 73.7
Llama 4 Behemoth
- 73.2
LongCat-Flash-Chat
- 73.2
InternVL3.5-241B-A28B
- 73.2
DiffusionGemma 26B A4B
- 73.0
Nemotron 3 Nano 30B-A3B
- 73.0
Step 3
- 73.0 Claude Haiku 4.5
- 72.6
Qwen3-Max
- 72.4 Beam 1.0 Base
- 72.2
Nemotron 3 Nano Omni 30B-A3B
- 72.0
Llama-3.3-Nemotron-Super-49B-v1.5
- 71.9
Solar Pro 3 (March 2026 update)
- 71.7
DeepSeek-R1
- 71.7
Gemini 2.5 Flash-Lite
- 71.5
seed-oss-36b-instruct
- 71.5 gpt-oss-20b
- 71.5
Magistral Medium 1.1
- 71.4 GPT-4.5
- 71.2
Hunyuan-A13B-Instruct
- 71.2 GPT-5 nano
- 71.2
Ministral 3 14B
- 71.2
Mistral Small 4
- 71.1
Qwen3-235B-A22B
- 70.8
STEP3-VL-10B
- 70.8
Magistral Medium 1.0
- 70.6 Hermes 4 405B
- 70.1
Qwen3-30B-A3B-Thinking
- 70.1
Magistral Small 1.2
- 70.0
MiniMax-M1-80k
- 69.8
Llama 4 Maverick
- 69.6
Qwen3-Omni-30B-A3B
- 69.3
Hunyuan-T1
- 69.3
Phi-4-Reasoning-plus
- 68.4
Qwen3-32B
- 68.2
Magistral Small 1.0
- 68.1
Solar Open 100B
- 67.6
DeepSeek-V3
- 67.1
Phi-4-reasoning
- 66.9 GPT-4.1
- 66.8 INTELLECT-2
- 66.7
Llama-3.3-Nemotron-Super-49B-v1
- 66.7
Gemini 2.5 Flash-Lite
- 66.7
Gemini 2.5 Pro
- 66.5 Sarvam 30B
- 66.4 Granite 4.2 30B
- 66.2
Nemotron 3.5 Super 120B-A12B Base
- 66.1
EXAONE Deep 32B
- 66.1 GLM-Z1-32B-0414
- 65.8 GPT-4.1 mini
- 65.8
Qwen3-30B-A3B
- 65.8
Magistral Small 1.1
- 65.7
Gemini 2.0 Pro
- 65.5 Hermes 4.3 36B
- 65.4
QWQ-Plus
- 65.3
QwQ-32B
- 65.2
DeepSeek-R1-Distill-Llama-70B
- 65.0 Claude 3.5 Sonnet (October 2024)
- 64.1
DeepSeek-R1-Distill-Qwen-32B
- 64.1
Gemini 2.0 Flash
- 64.1
Nemotron Nano 12B v2 VL
- 64.0
Qwen3-14B
- 64.0 nvidia-nemotron-nano-9b-v2
- 63.4 LFM3-1.7B
- 63.3
Trinity Large Preview
- 62.4 o1-mini
- 62.1
Gemini 2.0 Flash
- 62.0
Qwen3-8B
- 61.3
Falcon H1R 7B
- 61.1 Reka Flash 3
- 60.6
Cogito v1 Preview 70B
- 59.5
Mistral Medium 3
- 59.4 Claude 3.5 Sonnet
- 59.1
DeepSeek-V3
- 59.1
DeepSeek-R1-Distill-Qwen-14B
- 58.6
Gemma 4 E4B
- 58.5
Trinity Mini
- 58.1
Olmo 3 32B Think
- 57.5
Hunyuan-TurboS
- 57.5
Olmo 3.1 32B Think
- 57.2
Gemini 1.5 Pro
- 57.2
Llama 4 Scout
- 57.1
Amazon Nova Premier
- 56.1
Qwen2.5-Max
- 56.1
Phi-4
- 56.0 Grok-2
- 55.9
Qwen3-4B
- 55.6
Qwen3-30B-A3B-Instruct
- 54.4
MiniMax-Text-01
- 54.4 MiMo-7B-RL
- 53.8 Grok-2
- 53.6 GPT-4o
- 53.2
Nemotron 3 Nano 4B
- 52.0
Phi-4-mini-reasoning
- 51.6
Qwen3.5-2B
- 51.5
Gemini 2.0 Flash-Lite
- 51.3
Mistral Large 2
- 51.1
Llama 3.1-405B
- 51.0 Mercury
- 50.8
Cohere Command A
- 50.5
Llama 3.3 70B Instruct
- 50.4 Claude 3 Opus
- 50.3 GPT-4.1 nano
- 50.0
Nemotron 3 Ultra 550B-A55B Base
- 49.7
Falcon-H1-34B
- 49.5
Qwen2.5-32B
- 49.3 GPT-4 Turbo
- 49.2 GPT-4o
- 49.1
Qwen2.5-72B
- 49.0
Mistral Small 3.2
- 49.0
Mistral Large 2
- 49.0
Qwen2.5-VL-72B
- 48.1
Qwen Plus
- 48.0 Sarvam-M
- 47.9 GPT-4o
- 47.7
Gemma 3 27B
- 47.5
Mistral Small 3.1
- 47.4
Llama 3.3 70B
- 47.3
Gemini 1.5 Flash
- 47.3
Mistral Small 3
- 46.7
Llama 3.1-70B
- 46.3
Tulu 3 (Tülu 3) 70B
- 46.2
Gemini 1.5 Pro
- 45.8 qwen3-4b-instruct-2507
- 45.8 Granite 4.1 30B
- 45.5
Qwen2.5-14B
- 45.1 glm-4.7-flash_none
- 45.1
Phi-4-mini-flash-reasoning
- 43.9
Mistral Large 3
- 43.4
WizardLM-2 8x22B
- 43.4
Gemma 4 E2B
- 42.4
Qwen2-72B
- 42.4 GPT-4 Turbo
- 42.3
Qwen-Turbo
- 42.3 Kimi-VL-A3B-Thinking-2506
- 41.6 Claude 3.5 Haiku
- 41.5
Gemini 1.5 Pro
- 41.0
Llama 3.2 90B
- 40.9
Gemma 3 12B
- 40.6 Granite 4.0 H Small
- 40.6
Code World Model (CWM)
- 40.6 Claude 3 Sonnet
- 40.6
Llama 3-70B
- 40.4
Gemini 1.5 Flash
- 40.2 GPT-4o mini
- 40.1
Qwen3-1.7B
- 39.9
Qwen2.5-14B-Instruct-1M
- 38.9 LFM2.5-1.2B-Instruct
- 38.8
Mistral Large
- 37.9 LFM2.5-1.2B-Thinking
- 37.5 Hermes 2 Theta Llama-3 70B
- 37.4
InternLM3-8B-Instruct
- 36.8
Phi-3.5-MoE
- 36.5
Gemma 2 27B
- 36.4
Qwen2.5-7B
- 36.3 Claude 3 Haiku
- 35.9
OLMo 2 32B Instruct
- 35.7 GPT-4
- 35.7 GLM-4
- 34.7 Claude 2
- 34.3
phi-3-small 7.4B
- 34.2
Llama 3-8B
- 34.1
Mixtral 8x22B
- 34.0
Gemini 1.0 Pro
- 33.9 Eurus-2-7B-PRIME
- 33.8
DeepSeek-R1-Distill-Qwen-1.5B
- 33.3
Nemotron 3.5 Lightning 30B-A3B Base
- 33.0 Claude 2.1
- 33.0
Gemini 1.5 Flash 8B
- 32.9
DBRX
- 32.8
phi-3-mini 3.8B
- 32.8
Llama 3.1-8B
- 32.0
Yi-1.5-34B
- 31.5 LFM2-1.2B
- 30.8
Gemma 3 4B
- 30.8
Qwen2.5-Omni-7B
- 30.7
Qwen1.5-32B
- 30.6 GPT-4
- 30.6
Mixtral 8x7B
- 30.4
Phi-4 Mini
- 29.9 granite-4.0-1b
- 29.9
Mistral NeMo
- 29.3 LFM2-8B-A1B
- 28.8
Qwen1.5-72B
- 28.3 granite-4.0-micro
- 28.0 GPT-3.5 Turbo
- 27.8
AFM-4.5B
- 27.6
phi-3-medium 14B
- 27.5
Gemma 2 9B
- 27.2
Phi-3.5-mini
- 27.2 GPT-3.5 Turbo
- 27.1
Ministral 8B
- 26.8 ChatGLM3-6B
- 26.3
Llama 2-70B
- 25.3
Ministral 3B
- 24.6
DeepSeek LLM 67B
- 23.9
Llama 3.2 1B
- 23.7
Gemma 3n E4B
- 19.9
Gemma 3 1B
- 15.2
Mistral 7B v0.3
- 14.7
Yi-34B
- 11.9
Qwen3.5-0.8B
- 11.2 granite-4.0-350m
- 9.28 deepseek-r1-0528-qwen3-8b
- 1 GPT-6.2 Astra 97%
- 2
Gemini 4 Carbon 97%
- 3
Gemini 4 Hydrogen xhigh 96.8%
- 4
Gemini 4 Argon high 96.4%
- 5 Claude Fable 5.2 with fallback, max 96%
- 6 GPT-6 Astra 96%
- 7 GPT-6.2 Sol 96%
- 8 Claude Sonnet 5.5 max 95.6%
- 9
Gemini 3.8 Flash high 95.4%
- 10 GPT-6.1 Sol max 95.4%
- 11 Grok 4.6 95%
- 12
Gemini 3.7 Flash high 94.8%
- 13 GPT-5.4 Pro xhigh 94.6%
- 14 GPT-5.6 Sol 94.6%
- 15 Claude Mythos Preview 94.5%
- 16
Gemini 3.1 Pro high 94.4%
- 17 GPT-6 Sol max 94.3%
- 18 GPT-6.1 Terra max 94.3%
- 19 Claude Opus 4.7 94.2%
- 20 Kimi K3.1 Pro 94.2%
- 21
Gemini 3.6 Flash high 94.1%
- 22 Claude Opus 5.5 with fallback, max 94.1%
- 23 GPT-5.5 xhigh 94%
- 24 GPT-5.5 Pro xhigh 93.9%
- 25 Claude Fable 5 with fallback, max 93.9%
- 26 Claude Opus 5 max 93.9%
- 27 Claude Opus 4.8 93.6%
- 28 Kimi K3 93.5%
- 29
Step 5 Preview 93.5%
- 30 Grok 4.5 high 93.4%
- 31 GPT-5.6 Terra max 93.3%
- 32 GPT-5.4 xhigh 93.3%
- 33 GPT-5.2 Pro 93.2%
- 34
DeepSeek V4.1 Flash 1017 max 93.2%
- 35
Doubao Seed Evolving Nov 2026, high 93.1%
- 36
Inkling 2 Small 93%
- 37
Gemini 3.5 Flash high 92.8%
- 38 Grok 4.7 xhigh 92.7%
- 39
Qwen 3.8 Max xhigh 92.7%
- 40
Gemini 3 Pro 92.6%
- 41 GPT-5.2 92.4%
- 42
Qwen3.7-Max 92.4%
- 43 GPT-5.6 Luna 92.3%
- 44
Tencent Hy4 preview 92.3%
- 45
Qwen3.8 Max (0902) xhigh 92.3%
- 46 GLM-5.2 max 91.9%
- 47
Qwen3.8 Flash 91.7%
- 48
DeepSeek V4 Pro 0813 max 91.7%
- 49 Claude Opus 4.6 91.3%
- 50
DeepSeek V4 Flash 0731 max 91%
- 51
ERNIE 5.1 91%
- 52
Qwen3.8-Omni-Flash 91%
- 53
DeepSeek-V4-Pro high 90.9%
- 54
MiniMax-M3 90.9%
- 55 GLM-5.3 max 90.9%
- 56
DeepSeek V4.1 Flash 90.9%
- 57 Kimi K2.6 90.8%
- 58 Claude Sonnet 5 xhigh 90.5%
- 59 Beam 1.0 90.5%
- 60 GPT-6 Luna max 90.5%
- 61
Gemini 3 Flash 90.4%
- 62
Qwen 3.6 Plus 90.4%
- 63
Tencent Hy3 90.4%
- 64
Qwen3.7-Plus 90.3%
- 65 GLM-5.3-Flash max 90.2%
- 66
Mistral Large 4 high 90.2%
- 67 Claude Sonnet 4.6 89.9%
- 68 GLM-5.1 89.9%
- 69
Muse Spark 89.8%
- 70
MiniMax-M2.7 89.8%
- 71 Claude Haiku 5.5 max 89.7%
- 72
Doubao Seed 2.1 Turbo high 89.6%
- 73
Nemotron 3.5 Super VL 89.5%
- 74
Inkling-Small 89.5%
- 75 GPT-5 89.4%
- 76 Grok 4.20 89.3%
- 77
Doubao Seed 2.1 Pro high 89.3%
- 78 Mercury 3 89.2%
- 79
Qwen 3.8 27B 89.2%
- 80
Solar Pro 4 89%
- 81
Muse Spark Open 89%
- 82
Seed2.0 Pro 88.9%
- 83
LongCat-2.0 88.9%
- 84 Grok 4.3 Beta high 88.8%
- 85 Grok 4 Heavy 88.4%
- 86
Qwen3.5 397B-A17B 88.4%
- 87
Seed2.0 Lite (0428) 88.4%
- 88
Ring-2.6-1T 88.3%
- 89
Inkling xhigh 88.3%
- 90
Doubao-Seed-2.0 Pro high 88.2%
- 91
Doubao Seed 2.0 Lite Apr 2026, high 88.1%
- 92 GPT-5.1 88.1%
- 93
DeepSeek-V4-Flash 88.1%
- 94
Mistral Large 4 Preview high 88.1%
- 95 GPT-5.4 Mini 88%
- 96 Kimi K2.7 Code 87.9%
- 97 GLM-5 87.8%
- 98
Qwen3.6 27B 87.8%
- 99 Kimi K2.5 Fireworks 87.6%
- 100 Grok 4 87.5%
- 101
Qwen3-Max-2026-01-23 87.4%
- 102
Qwen 3.6 Max (Preview) 87.4%
- 103
Tencent Hy3 preview 87.2%
- 104 Claude Opus 4.5 87%
- 105
Nemotron 3 Ultra 87%
- 106
Gemini 3.1 Flash-Lite 86.9%
- 107
Qwen3.5-122B-A10B 86.6%
- 108
Inkling 2 Mini 86.6%
- 109
Nemotron 3.5 Bolt 74B-A9B 86.4%
- 110
Gemini 2.5 Pro Jun 2025 86.4%
- 111
ERNIE 5.0 86.4%
- 112
Solar Open 2 86.3%
- 113
Qwen 3.6 35B-A3B 86%
- 114 Grok 4 Fast 85.7%
- 115
DeepSeek-V3.2-Speciale 85.7%
- 116 GLM-4.7 85.7%
- 117 GPT-5.5 Instant 85.6%
- 118
Qwen3.5-27B 85.5%
- 119
LongCat-Flash-Thinking-2601 85.2%
- 120
MiniMax-M2.5 85.2%
- 121
Qwen 3.5 Plus (hosted 397B-A17B) 84.8%
- 122 Claude 3.7 Sonnet 84.8%
- 123 Grok 3 84.6%
- 124
MAI-Code-1-Flash 84.6%
- 125 Kimi K2 Thinking 84.5%
- 126 MiMo-V2-Flash 84.3%
- 127
Gemma 4 31B IT 84.3%
- 128
Qwen3.5-35B-A3B 84.2%
- 129
MAI-Thinking-1 84.2%
- 130 Grok-3 mini 84%
- 131
Gemini 2.5 Pro Mar 2025 84%
- 132 o3-pro 84%
- 133
Qwen3.5-Omni-Plus 83.9%
- 134 Claude Sonnet 4 83.8%
- 135
Seed1.8 83.8%
- 136
Step 3.5 Flash 83.5%
- 137
Muse Glimmer 83.5%
- 138
DeepSeek-V3.2 Thinking 83.4%
- 139 Claude Sonnet 4.5 83.4%
- 140
Qwen 3.6 Flash 2026-04-16 83.3%
- 141
Gemini 3.5 Flash-Lite high 83.3%
- 142 o3 83.3%
- 143 Claude Opus 4 83.3%
- 144
gemini-robotics-er-1.5-preview 83.3%
- 145
Gemma 4 26B A4B 83.2%
- 146
MiniMax-M2.1 83%
- 147
Gemini 2.5 Flash Jun 2025 82.8%
- 148 GPT-5.4 Nano 82.8%
- 149
Qwen 3.5 Flash (hosted 35B-A3B) 82.3%
- 150
Qwen3.7 Flash 82.3%
- 151 GPT-5 mini 82.3%
- 152
K-EXAONE 2.0 82.2%
- 153
Qwen3.5-9B 81.7%
- 154
Seed1.6-Thinking 81.5%
- 155
LongCat-Flash-Thinking 81.5%
- 156 o4-mini 81.4%
- 157
Nova 2.0 Pro Preview 81.4%
- 158
Qwen3-235B-A22B-Thinking Jul 2025 81.1%
- 159
DeepSeek-R1 May 2025 81%
- 160 GLM-4.6 81%
- 161 Claude Opus 4.1 80.9%
- 162
Gemini 2.5 Flash Sep 2025 80.8%
- 163
DeepSeek-V3.1-Terminus 80.7%
- 164
EXAONE 4.5 33B 80.5%
- 165 gpt-oss-120b 80.1%
- 166
DeepSeek-V3.1 80.1%
- 167
DeepSeek-V3.2-Exp 79.9%
- 168 o3-mini 79.7%
- 169
Nova 2 Lite 79.6%
- 170
Nemotron 3 Super 79.2%
- 171 GLM-4.5 79.1%
- 172
K-EXAONE 79.1%
- 173 Mercury 2.5 79%
- 174
Gemma 4 12B 78.8%
- 175 Sarvam 105B 78.7%
- 176
Ring-1T 78.6%
- 177 o1-preview 78.3%
- 178
Gemini 2.5 Flash Apr 2025 78.3%
- 179
MiniMax-M2 78%
- 180
Cogito v2.1 77.7%
- 181
Doubao-Seed-1.6 thinking 77.6%
- 182
Seed-Thinking-v1.5 77.3%
- 183
Intern-S1 77.3%
- 184
Qwen3-Next-80B-A3B 77.2%
- 185
Qwen3-VL-235B-A22B 77.1%
- 186
Spark-X2.6-4B 77.1%
- 187 o1 high 76.8%
- 188
Nemotron 3.5 Lightning 30B-A3B 76.5%
- 189
Trinity Large Thinking 76.3%
- 190
Magistral Medium 1.2 76.3%
- 191
Qwen3.5-4B 76.2%
- 192
Ling-2.6-1T 76.2%
- 193
Nemotron-Cascade 2 30B-A3B 76.1%
- 194
Cogito v2 Preview 671B MoE 76%
- 195
Llama-3.1-Nemotron-Ultra-253B-v1 76%
- 196
EXAONE 4.0 32B 75.4%
- 197 glm-4.7-flash 75.2%
- 198 Kimi K2 Jul 2025 75.1%
- 199 GLM-4.5-Air 75%
- 200
Qwen3-Coder-Next 74.5%
- 201 LFM2.5-24B-A2B 74.4%
- 202
LongCat-Flash-Omni 74.4%
- 203 INTELLECT-3 74.4%
- 204
Gemini 2.0 Flash Thinking Jan 2025 74.2%
- 205 Mercury 2 74%
- 206
Llama 4 Behemoth 73.7%
- 207
LongCat-Flash-Chat 73.2%
- 208
InternVL3.5-241B-A28B 73.2%
- 209
DiffusionGemma 26B A4B 73.2%
- 210
Nemotron 3 Nano 30B-A3B 73%
- 211
Step 3 73%
- 212 Claude Haiku 4.5 73%
- 213
Qwen3-Max 72.6%
- 214 Beam 1.0 Base 72.4%
- 215
Nemotron 3 Nano Omni 30B-A3B 72.2%
- 216
Llama-3.3-Nemotron-Super-49B-v1.5 72%
- 217
Solar Pro 3 (March 2026 update) 71.9%
- 218
DeepSeek-R1 71.7%
- 219
Gemini 2.5 Flash-Lite Sep 2025 71.7%
- 220
seed-oss-36b-instruct 71.5%
- 221 gpt-oss-20b 71.5%
- 222
Magistral Medium 1.1 71.5%
- 223 GPT-4.5 71.4%
- 224
Hunyuan-A13B-Instruct 71.2%
- 225 GPT-5 nano 71.2%
- 226
Ministral 3 14B 71.2%
- 227
Mistral Small 4 71.2%
- 228
Qwen3-235B-A22B 71.1%
- 229
STEP3-VL-10B 70.8%
- 230
Magistral Medium 1.0 70.8%
- 231 Hermes 4 405B 70.6%
- 232
Qwen3-30B-A3B-Thinking Jul 2025 70.1%
- 233
Magistral Small 1.2 70.1%
- 234
MiniMax-M1-80k 70%
- 235
Llama 4 Maverick 69.8%
- 236
Qwen3-Omni-30B-A3B 69.6%
- 237
Hunyuan-T1 69.3%
- 238
Phi-4-Reasoning-plus 69.3%
- 239
Qwen3-32B 68.4%
- 240
Magistral Small 1.0 68.2%
- 241
Solar Open 100B 68.1%
- 242
DeepSeek-V3 Mar 2025 67.6%
- 243
Phi-4-reasoning 67.1%
- 244 GPT-4.1 66.9%
- 245 INTELLECT-2 66.8%
- 246
Llama-3.3-Nemotron-Super-49B-v1 66.7%
- 247
Gemini 2.5 Flash-Lite Jun 2025 66.7%
- 248
Gemini 2.5 Pro May 2025 66.7%
- 249 Sarvam 30B 66.5%
- 250 Granite 4.2 30B 66.4%
- 251
Nemotron 3.5 Super 120B-A12B Base 66.2%
- 252
EXAONE Deep 32B 66.1%
- 253 GLM-Z1-32B-0414 66.1%
- 254 GPT-4.1 mini 65.8%
- 255
Qwen3-30B-A3B 65.8%
- 256
Magistral Small 1.1 65.8%
- 257
Gemini 2.0 Pro 65.7%
- 258 Hermes 4.3 36B 65.5%
- 259
QWQ-Plus 65.4%
- 260
QwQ-32B 65.3%
- 261
DeepSeek-R1-Distill-Llama-70B 65.2%
- 262 Claude 3.5 Sonnet (October 2024) 65%
- 263
DeepSeek-R1-Distill-Qwen-32B 64.1%
- 264
Gemini 2.0 Flash Feb 2025 64.1%
- 265
Nemotron Nano 12B v2 VL 64.1%
- 266
Qwen3-14B 64%
- 267 nvidia-nemotron-nano-9b-v2 64%
- 268 LFM3-1.7B 63.4%
- 269
Trinity Large Preview 63.3%
- 270 o1-mini high 62.4%
- 271
Gemini 2.0 Flash Dec 2024 62.1%
- 272
Qwen3-8B 62%
- 273
Falcon H1R 7B 61.3%
- 274 Reka Flash 3 61.1%
- 275
Cogito v1 Preview 70B 60.6%
- 276
Mistral Medium 3 59.5%
- 277 Claude 3.5 Sonnet 59.4%
- 278
DeepSeek-V3 59.1%
- 279
DeepSeek-R1-Distill-Qwen-14B 59.1%
- 280
Gemma 4 E4B 58.6%
- 281
Trinity Mini 58.5%
- 282
Olmo 3 32B Think 58.1%
- 283
Hunyuan-TurboS 57.5%
- 284
Olmo 3.1 32B Think 57.5%
- 285
Gemini 1.5 Pro Sep 2024 57.2%
- 286
Llama 4 Scout 57.2%
- 287
Amazon Nova Premier 57.1%
- 288
Qwen2.5-Max 56.1%
- 289
Phi-4 56.1%
- 290 Grok-2 Aug 2024 56%
- 291
Qwen3-4B 55.9%
- 292
Qwen3-30B-A3B-Instruct Jul 2025 55.6%
- 293
MiniMax-Text-01 54.4%
- 294 MiMo-7B-RL 54.4%
- 295 Grok-2 Dec 2024 53.8%
- 296 GPT-4o May 2024 53.6%
- 297
Nemotron 3 Nano 4B 53.2%
- 298
Phi-4-mini-reasoning 52%
- 299
Qwen3.5-2B 51.6%
- 300
Gemini 2.0 Flash-Lite 51.5%
- 301
Mistral Large 2 Nov 2024 51.3%
- 302
Llama 3.1-405B 51.1%
- 303 Mercury 51%
- 304
Cohere Command A 50.8%
- 305
Llama 3.3 70B Instruct 50.5%
- 306 Claude 3 Opus 50.4%
- 307 GPT-4.1 nano 50.3%
- 308
Nemotron 3 Ultra 550B-A55B Base 50%
- 309
Falcon-H1-34B 49.7%
- 310
Qwen2.5-32B 49.5%
- 311 GPT-4 Turbo Apr 2024 49.3%
- 312 GPT-4o Aug 2024 49.2%
- 313
Qwen2.5-72B 49.1%
- 314
Mistral Small 3.2 49%
- 315
Mistral Large 2 Jul 2024 49%
- 316
Qwen2.5-VL-72B 49%
- 317
Qwen Plus Jan 2025 48.1%
- 318 Sarvam-M 48%
- 319 GPT-4o Nov 2024 47.9%
- 320
Gemma 3 27B 47.7%
- 321
Mistral Small 3.1 47.5%
- 322
Llama 3.3 70B 47.4%
- 323
Gemini 1.5 Flash Sep 2024 47.3%
- 324
Mistral Small 3 47.3%
- 325
Llama 3.1-70B 46.7%
- 326
Tulu 3 (Tülu 3) 70B 46.3%
- 327
Gemini 1.5 Pro May 2024 46.2%
- 328 qwen3-4b-instruct-2507 45.8%
- 329 Granite 4.1 30B 45.8%
- 330
Qwen2.5-14B 45.5%
- 331 glm-4.7-flash_none none 45.1%
- 332
Phi-4-mini-flash-reasoning 45.1%
- 333
Mistral Large 3 43.9%
- 334
WizardLM-2 8x22B 43.4%
- 335
Gemma 4 E2B 43.4%
- 336
Qwen2-72B 42.4%
- 337 GPT-4 Turbo Nov 2023 42.4%
- 338
Qwen-Turbo 42.3%
- 339 Kimi-VL-A3B-Thinking-2506 42.3%
- 340 Claude 3.5 Haiku 41.6%
- 341
Gemini 1.5 Pro Feb 2024 41.5%
- 342
Llama 3.2 90B 41%
- 343
Gemma 3 12B 40.9%
- 344 Granite 4.0 H Small 40.6%
- 345
Code World Model (CWM) 40.6%
- 346 Claude 3 Sonnet 40.6%
- 347
Llama 3-70B 40.6%
- 348
Gemini 1.5 Flash May 2024 40.4%
- 349 GPT-4o mini 40.2%
- 350
Qwen3-1.7B 40.1%
- 351
Qwen2.5-14B-Instruct-1M 39.9%
- 352 LFM2.5-1.2B-Instruct 38.9%
- 353
Mistral Large 38.8%
- 354 LFM2.5-1.2B-Thinking 37.9%
- 355 Hermes 2 Theta Llama-3 70B 37.5%
- 356
InternLM3-8B-Instruct 37.4%
- 357
Phi-3.5-MoE 36.8%
- 358
Gemma 2 27B 36.5%
- 359
Qwen2.5-7B 36.4%
- 360 Claude 3 Haiku 36.3%
- 361
OLMo 2 32B Instruct 35.9%
- 362 GPT-4 Mar 2023 35.7%
- 363 GLM-4 35.7%
- 364 Claude 2 34.7%
- 365
phi-3-small 7.4B 34.3%
- 366
Llama 3-8B 34.2%
- 367
Mixtral 8x22B 34.1%
- 368
Gemini 1.0 Pro 34%
- 369 Eurus-2-7B-PRIME 33.9%
- 370
DeepSeek-R1-Distill-Qwen-1.5B 33.8%
- 371
Nemotron 3.5 Lightning 30B-A3B Base 33.3%
- 372 Claude 2.1 33%
- 373
Gemini 1.5 Flash 8B 33%
- 374
DBRX 32.9%
- 375
phi-3-mini 3.8B 32.8%
- 376
Llama 3.1-8B 32.8%
- 377
Yi-1.5-34B 32%
- 378 LFM2-1.2B 31.5%
- 379
Gemma 3 4B 30.8%
- 380
Qwen2.5-Omni-7B 30.8%
- 381
Qwen1.5-32B 30.7%
- 382 GPT-4 Jun 2023 30.6%
- 383
Mixtral 8x7B 30.6%
- 384
Phi-4 Mini 30.4%
- 385 granite-4.0-1b 29.9%
- 386
Mistral NeMo 29.9%
- 387 LFM2-8B-A1B 29.3%
- 388
Qwen1.5-72B 28.8%
- 389 granite-4.0-micro 28.3%
- 390 GPT-3.5 Turbo Nov 2023 28%
- 391
AFM-4.5B 27.8%
- 392
phi-3-medium 14B 27.6%
- 393
Gemma 2 9B 27.5%
- 394
Phi-3.5-mini 27.2%
- 395 GPT-3.5 Turbo Jan 2024 27.2%
- 396
Ministral 8B 27.1%
- 397 ChatGLM3-6B 26.8%
- 398
Llama 2-70B 26.3%
- 399
Ministral 3B 25.3%
- 400
DeepSeek LLM 67B 24.6%
- 401
Llama 3.2 1B 23.9%
- 402
Gemma 3n E4B 23.7%
- 403
Gemma 3 1B 19.9%
- 404
Mistral 7B v0.3 15.2%
- 405
Yi-34B 14.7%
- 406
Qwen3.5-0.8B 11.9%
- 407 granite-4.0-350m 11.2%
- 408 deepseek-r1-0528-qwen3-8b 9.28%
Sources: Labs’ model cards, official GPQA Diamond leaderboard and Epoch AI runs (CC BY 4.0)
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | | 97% | Oct 2026 | Nov 2026 | Official leaderboard | |
| 2 | Grok 4.6 | 95% | Aug 2026 | Oct 2026 | Official leaderboard | |
| 3 | GPT-6 Astra | 95% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 4 | | 95% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 5 | | 93.9% | Feb 2026 | Oct 2026 | Official leaderboard | |
| 6 | GPT-6.1 Sol | 93.9% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 7 | | 92.9% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 8 | Claude Opus 5 | 92.9% | Jul 2026 | Oct 2026 | Official leaderboard | |
| 9 | Grok 4.7 | 92.6% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 10 | GPT-6 Sol | 91.9% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 11 | Claude Sonnet 5.5 | 91.9% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 12 | Claude Opus 5.5 | 90.8% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 13 | | 89.9% | Jul 2026 | Oct 2026 | Official leaderboard | |
| 14 | | 89.4% | Oct 2026 | Nov 2026 | Official leaderboard | |
| 15 | Claude Haiku 5.5 | 84.8% | Oct 2026 | Oct 2026 | Official leaderboard | |
| 16 | | 83.2% | Apr 2026 | Oct 2026 | Official leaderboard | |
| 17 | | 82.8% | Apr 2026 | Oct 2026 | Official leaderboard | |
| 18 | GPT-6 Luna | 81.8% | Sep 2026 | Oct 2026 | Official leaderboard | |
| 19 | GPT-5.6 Luna | 80.8% | Jul 2026 | Oct 2026 | Official leaderboard | |
| 20 | GPT-5.6 Terra | 79.3% | Jul 2026 | Oct 2026 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | 97% | Oct 2026 | Official leaderboard | ||
| 2 | GPT-6.2 Astra | 96.8% | max | Nov 2026 | Epoch AI run |
| 3 | 96.8% | xhigh | Nov 2026 | Epoch AI run | |
| 4 | 96.4% | high | Oct 2026 | Epoch AI run | |
| 5 | Claude Fable 5.2 | 96% | with fallback, max | Nov 2026 | Epoch AI run |
| 6 | GPT-6 Astra | 95.8% | max | Sep 2026 | Epoch AI run |
| 7 | Claude Sonnet 5.5 | 95.6% | max | Sep 2026 | Epoch AI run |
| 8 | 95.4% | high | Sep 2026 | Epoch AI run | |
| 9 | GPT-6.1 Sol | 95.4% | max | Sep 2026 | Epoch AI run |
| 10 | Grok 4.6 | 95% | Aug 2026 | Official leaderboard | |
| 11 | 94.8% | high | Aug 2026 | Epoch AI run | |
| 12 | GPT-5.4 Pro | 94.6% | xhigh | Mar 2026 | Epoch AI run |
| 13 | 94.4% | high | Feb 2026 | Epoch AI run | |
| 14 | GPT-6 Sol | 94.3% | max | Sep 2026 | Epoch AI run |
| 15 | GPT-6.1 Terra | 94.3% | max | Oct 2026 | Epoch AI run |
| 16 | Kimi K3.1 Pro | 94.2% | Together, max | Oct 2026 | Epoch AI run |
| 17 | 94.1% | high | Jul 2026 | Epoch AI run | |
| 18 | Claude Opus 5.5 | 94.1% | with fallback, max | Sep 2026 | Epoch AI run |
| 19 | GPT-5.5 | 94% | xhigh | Apr 2026 | Epoch AI run |
| 20 | GPT-5.5 Pro | 93.9% | xhigh | Apr 2026 | Epoch AI run |
| 21 | Claude Fable 5 | 93.9% | with fallback, max | Jun 2026 | Epoch AI run |
| 22 | Claude Opus 5 | 93.9% | max | Jul 2026 | Epoch AI run |
| 23 | GPT-5.6 Sol | 93.5% | max | Jul 2026 | Epoch AI run |
| 24 | Grok 4.5 | 93.4% | high | Jul 2026 | Epoch AI run |
| 25 | GPT-5.6 Terra | 93.3% | max | Jul 2026 | Epoch AI run |
| 26 | GPT-5.4 | 93.3% | xhigh | Mar 2026 | Epoch AI run |
| 27 | 93.2% | max | Oct 2026 | Epoch AI run | |
| 28 | Kimi K3 | 93.1% | max | Jul 2026 | Epoch AI run |
| 29 | 93.1% | high | Nov 2026 | Epoch AI run | |
| 30 | 92.8% | high | May 2026 | Epoch AI run | |
| 31 | Grok 4.7 | 92.7% | xhigh | Sep 2026 | Epoch AI run |
| 32 | 92.7% | xhigh | Aug 2026 | Epoch AI run | |
| 33 | 92.6% | Nov 2025 | Epoch AI run | ||
| 34 | 92.3% | xhigh | Sep 2026 | Epoch AI run | |
| 35 | GLM-5.2 | 91.9% | max | Jun 2026 | Epoch AI run |
| 36 | 91.7% | max | Aug 2026 | Epoch AI run | |
| 37 | GPT-5.6 Luna | 91.6% | max | Jul 2026 | Epoch AI run |
| 38 | GPT-5.2 | 91.4% | xhigh | Dec 2025 | Epoch AI run |
| 39 | Claude Opus 4.8 | 91% | max | May 2026 | Epoch AI run |
| 40 | 91% | max | Jul 2026 | Epoch AI run | |
| 41 | 90.9% | high | Apr 2026 | Epoch AI run | |
| 42 | 90.9% | May 2026 | Epoch AI run | ||
| 43 | 90.9% | Jun 2026 | Epoch AI run | ||
| 44 | GLM-5.3 | 90.9% | max | Aug 2026 | Epoch AI run |
| 45 | Kimi K2.6 | 90.8% | Apr 2026 | Epoch AI run | |
| 46 | Claude Opus 4.6 | 90.5% | 32k thinking | Feb 2026 | Epoch AI run |
| 47 | Claude Sonnet 5 | 90.5% | xhigh | Jun 2026 | Epoch AI run |
| 48 | GPT-6 Luna | 90.5% | max | Sep 2026 | Epoch AI run |
| 49 | Claude Opus 4.7 | 90.2% | xhigh | Apr 2026 | Epoch AI run |
| 50 | GLM-5.3-Flash | 90.2% | max | Aug 2026 | Epoch AI run |
| 51 | 90.2% | high | Oct 2026 | Epoch AI run | |
| 52 | GLM-5.1 | 89.9% | Apr 2026 | Epoch AI run | |
| 53 | 89.8% | max | Sep 2026 | Epoch AI run | |
| 54 | 89.8% | Apr 2026 | Epoch AI run | ||
| 55 | Claude Haiku 5.5 | 89.7% | max | Oct 2026 | Epoch AI run |
| 56 | 89.6% | high | Jun 2026 | Epoch AI run | |
| 57 | 89.4% | high | Dec 2025 | Epoch AI run | |
| 58 | Grok 4.20 | 89.3% | Feb 2026 | Epoch AI run | |
| 59 | 89.3% | high | Jun 2026 | Epoch AI run | |
| 60 | Grok 4.3 Beta | 88.8% | high | Apr 2026 | Epoch AI run |
| 61 | 88.6% | xhigh | Oct 2026 | Epoch AI run | |
| 62 | 88.5% | xhigh | Jul 2026 | Epoch AI run | |
| 63 | 88.4% | 2026-04-02 | Mar 2026 | Epoch AI run | |
| 64 | 88.3% | xhigh | Jul 2026 | Epoch AI run | |
| 65 | 88.2% | high | Feb 2026 | Epoch AI run | |
| 66 | 88.1% | high | Apr 2026 | Epoch AI run | |
| 67 | 88.1% | high | Oct 2026 | Epoch AI run | |
| 68 | 87.9% | Jun 2026 | Epoch AI run | ||
| 69 | Kimi K2.7 Code | 87.9% | Jun 2026 | Epoch AI run | |
| 70 | GLM-5 | 87.8% | Feb 2026 | Epoch AI run | |
| 71 | GPT-5.1 | 87.6% | high | Nov 2025 | Epoch AI run |
| 72 | Kimi K2.5 | 87.6% | Fireworks | Jan 2026 | Epoch AI run |
| 73 | Claude Sonnet 4.6 | 87.4% | 32k thinking | Feb 2026 | Epoch AI run |
| 74 | 87.4% | Apr 2026 | Epoch AI run | ||
| 75 | Grok 4 | 87% | Jul 2025 | Epoch AI run | |
| 76 | GPT-5.4 Mini | 86.9% | xhigh | Mar 2026 | Epoch AI run |
| 77 | 86.4% | none | Feb 2026 | Epoch AI run | |
| 78 | GPT-5 | 86.2% | high | Aug 2025 | Epoch AI run |
| 79 | Claude Opus 4.5 | 86% | 32k thinking | Nov 2025 | Epoch AI run |
| 80 | 85.9% | Apr 2026 | Epoch AI run | ||
| 81 | 85.3% | Jun 2026 | Epoch AI run | ||
| 82 | 85.3% | Jun 2025 | Epoch AI run | ||
| 83 | 84.8% | Feb 2026 | Epoch AI run | ||
| 84 | 84.8% | none | Apr 2026 | Epoch AI run | |
| 85 | Kimi K2 Thinking | 84.2% | Nov 2025 | Epoch AI run | |
| 86 | 83.8% | Mar 2025 | Epoch AI run | ||
| 87 | 83.5% | Feb 2026 | Epoch AI run | ||
| 88 | 83.4% | Thinking | Dec 2025 | Epoch AI run | |
| 89 | GLM-4.7 | 83.3% | Dec 2025 | Epoch AI run | |
| 90 | 83.3% | 2026-04-16 | Apr 2026 | Epoch AI run | |
| 91 | 83.3% | high | Jul 2026 | Epoch AI run | |
| 92 | 83.2% | Apr 2026 | Official leaderboard | ||
| 93 | 82.8% | Apr 2026 | Official leaderboard | ||
| 94 | GPT-5.5 Instant | 82.5% | May 2026 | Epoch AI run | |
| 95 | Claude Sonnet 4.5 | 82.3% | 59k thinking | Sep 2025 | Epoch AI run |
| 96 | 82.3% | Feb 2026 | Epoch AI run | ||
| 97 | 82.3% | Jul 2026 | Epoch AI run | ||
| 98 | o3 | 81.8% | high | Apr 2025 | Epoch AI run |
| 99 | 81.8% | high | Mar 2026 | Epoch AI run | |
| 100 | 80% | Jul 2025 | Epoch AI run | ||
| 101 | Claude 3.7 Sonnet | 79.7% | 64k thinking | Feb 2025 | Epoch AI run |
| 102 | o4-mini | 79.6% | high | Apr 2025 | Epoch AI run |
| 103 | Claude Sonnet 4 | 79.2% | 59K | May 2025 | Epoch AI run |
| 104 | 79% | Feb 2026 | Epoch AI run | ||
| 105 | GPT-5.4 Nano | 78.5% | high | Mar 2026 | Epoch AI run |
| 106 | 77.6% | thinking | Jun 2025 | Epoch AI run | |
| 107 | Claude Opus 4.1 | 77.3% | 16K | Aug 2025 | Epoch AI run |
| 108 | o3-mini | 77% | high | Jan 2025 | Epoch AI run |
| 109 | o1 | 76.8% | high | Dec 2024 | Epoch AI run |
| 110 | 76.3% | May 2025 | Epoch AI run | ||
| 111 | Grok-3 mini | 76.3% | low | Apr 2025 | Epoch AI run |
| 112 | Claude Opus 4 | 76.3% | 16K | May 2025 | Epoch AI run |
| 113 | Grok 3 | 75.8% | Apr 2025 | Epoch AI run | |
| 114 | gpt-oss-120b | 75.8% | high | Aug 2025 | Epoch AI run |
| 115 | GPT-5 mini | 75% | high | Aug 2025 | Epoch AI run |
| 116 | 72.6% | Sep 2025 | Epoch AI run | ||
| 117 | 71.7% | Jan 2025 | Epoch AI run | ||
| 118 | 71.5% | Aug 2025 | Epoch AI run | ||
| 119 | Claude Haiku 4.5 | 71.2% | 32K | Oct 2025 | Epoch AI run |
| 120 | 70.7% | Apr 2025 | Epoch AI run | ||
| 121 | 70.1% | Jul 2025 | Epoch AI run | ||
| 122 | GPT-5 nano | 69.4% | high | Aug 2025 | Epoch AI run |
| 123 | GPT-4.5 | 68.7% | Feb 2025 | Epoch AI run | |
| 124 | 67.6% | Mar 2025 | Epoch AI run | ||
| 125 | 67% | FP8 | Apr 2025 | Epoch AI run | |
| 126 | GPT-4.1 | 66.9% | Apr 2025 | Epoch AI run | |
| 127 | 66.7% | May 2025 | Epoch AI run | ||
| 128 | GPT-4.1 mini | 65.8% | Apr 2025 | Epoch AI run | |
| 129 | 65.7% | Apr 2025 | Epoch AI run | ||
| 130 | 65.7% | Feb 2025 | Epoch AI run | ||
| 131 | 65.4% | Apr 2025 | Epoch AI run | ||
| 132 | 65.3% | Mar 2025 | Epoch AI run | ||
| 133 | 64.1% | Jan 2025 | Epoch AI run | ||
| 134 | 64.1% | Feb 2025 | Epoch AI run | ||
| 135 | 63.8% | Apr 2025 | Epoch AI run | ||
| 136 | o1-mini | 62.4% | high | Sep 2024 | Epoch AI run |
| 137 | 61.7% | Apr 2025 | Epoch AI run | ||
| 138 | gpt-oss-20b | 60.8% | medium | Aug 2025 | Epoch AI run |
| 139 | glm-4.7-flash | 60.5% | Jan 2026 | Epoch AI run | |
| 140 | 59.5% | May 2025 | Epoch AI run | ||
| 141 | 57.2% | Sep 2024 | Epoch AI run | ||
| 142 | 57.1% | Jan 2025 | Epoch AI run | ||
| 143 | 56.8% | Apr 2025 | Epoch AI run | ||
| 144 | 56.5% | Dec 2024 | Epoch AI run | ||
| 145 | 56.1% | Jan 2025 | Epoch AI run | ||
| 146 | 56.1% | Dec 2024 | Epoch AI run | ||
| 147 | 56.1% | Jun 2025 | Epoch AI run | ||
| 148 | 55.7% | Jan 2025 | Epoch AI run | ||
| 149 | 55.6% | Jul 2025 | Epoch AI run | ||
| 150 | Claude 3.5 Sonnet (October 2024) | 55.3% | Oct 2024 | Epoch AI run | |
| 151 | Claude 3.5 Sonnet | 54% | Jun 2024 | Jun 2024 | Epoch AI run |
| 152 | Grok-2 Dec 2024 | 53.8% | Dec 2024 | Epoch AI run | |
| 153 | 52.3% | Apr 2025 | Epoch AI run | ||
| 154 | 51.8% | Apr 2025 | Epoch AI run | ||
| 155 | 51.3% | Nov 2024 | Epoch AI run | ||
| 156 | 50.9% | Jul 2024 | Epoch AI run | ||
| 157 | o1-preview | 50.3% | Sep 2024 | Epoch AI run | |
| 158 | GPT-4o Aug 2024 | 49.2% | Aug 2024 | Epoch AI run | |
| 159 | 49.1% | Sep 2024 | Epoch AI run | ||
| 160 | 49% | Jun 2025 | Epoch AI run | ||
| 161 | 49% | Jul 2024 | Epoch AI run | ||
| 162 | GPT-4.1 nano | 48.9% | Apr 2025 | Epoch AI run | |
| 163 | GPT-4o May 2024 | 48.9% | May 2024 | Epoch AI run | |
| 164 | 48.1% | Jan 2025 | Epoch AI run | ||
| 165 | GPT-4o Nov 2024 | 47.9% | Nov 2024 | Epoch AI run | |
| 166 | 47.7% | Mar 2025 | Epoch AI run | ||
| 167 | 47.6% | Sep 2025 | Epoch AI run | ||
| 168 | 47.5% | Mar 2025 | Epoch AI run | ||
| 169 | 47.4% | Dec 2024 | Epoch AI run | ||
| 170 | 47.3% | Sep 2024 | Epoch AI run | ||
| 171 | 47.3% | Jan 2025 | Epoch AI run | ||
| 172 | Claude 3 Opus | 47.2% | Feb 2024 | Epoch AI run | |
| 173 | GPT-4 Turbo Apr 2024 | 46.6% | Apr 2024 | Epoch AI run | |
| 174 | 46.3% | Nov 2024 | Epoch AI run | ||
| 175 | 46.1% | Sep 2024 | Epoch AI run | ||
| 176 | 45.9% | May 2024 | Epoch AI run | ||
| 177 | qwen3-4b-instruct-2507 | 45.8% | Aug 2025 | Epoch AI run | |
| 178 | glm-4.7-flash_none | 45.1% | none | Epoch AI run | |
| 179 | 44.7% | Jan 2025 | Epoch AI run | ||
| 180 | 44.2% | Jul 2024 | Epoch AI run | ||
| 181 | 43.4% | Apr 2024 | Epoch AI run | ||
| 182 | GPT-4 Turbo Nov 2023 | 42.4% | Nov 2023 | Epoch AI run | |
| 183 | 41.8% | Nov 2024 | Epoch AI run | ||
| 184 | 41% | Sep 2024 | Epoch AI run | ||
| 185 | 40.8% | Jun 2024 | Epoch AI run | ||
| 186 | Claude 3 Sonnet | 40.6% | Feb 2024 | Epoch AI run | |
| 187 | 40.6% | Apr 2024 | Epoch AI run | ||
| 188 | 40.4% | May 2024 | Epoch AI run | ||
| 189 | 39.5% | Mar 2025 | Epoch AI run | ||
| 190 | 38.8% | Feb 2024 | Epoch AI run | ||
| 191 | Claude 3.5 Haiku | 38.1% | Oct 2024 | Oct 2024 | Epoch AI run |
| 192 | 38% | Apr 2025 | Epoch AI run | ||
| 193 | GPT-4o mini | 37.7% | Jul 2024 | Epoch AI run | |
| 194 | Hermes 2 Theta Llama-3 70B | 37.5% | Jun 2024 | Epoch AI run | |
| 195 | 36.5% | Jun 2024 | Epoch AI run | ||
| 196 | Claude 3 Haiku | 36.3% | Mar 2024 | Epoch AI run | |
| 197 | GPT-4 Mar 2023 | 35.7% | Mar 2023 | Epoch AI run | |
| 198 | 35.5% | Sep 2024 | Epoch AI run | ||
| 199 | Claude 2 | 34.7% | Jul 2023 | Epoch AI run | |
| 200 | 34.1% | Apr 2024 | Epoch AI run | ||
| 201 | 34% | Dec 2023 | Epoch AI run | ||
| 202 | Eurus-2-7B-PRIME | 33.9% | Dec 2024 | Epoch AI run | |
| 203 | 33.6% | Jan 2025 | Epoch AI run | ||
| 204 | Claude 2.1 | 33% | Nov 2023 | Epoch AI run | |
| 205 | 33% | Oct 2024 | Epoch AI run | ||
| 206 | 32.9% | Mar 2024 | Epoch AI run | ||
| 207 | 32% | May 2024 | Epoch AI run | ||
| 208 | 30.7% | Apr 2024 | Epoch AI run | ||
| 209 | GPT-4 Jun 2023 | 30.6% | Jun 2023 | Epoch AI run | |
| 210 | 30.6% | Dec 2023 | Epoch AI run | ||
| 211 | 29.9% | Jul 2024 | Epoch AI run | ||
| 212 | 28.8% | Feb 2024 | Epoch AI run | ||
| 213 | granite-4.0-micro | 28.3% | Oct 2025 | Epoch AI run | |
| 214 | GPT-3.5 Turbo Nov 2023 | 28% | Nov 2023 | Epoch AI run | |
| 215 | 27.6% | Apr 2024 | Epoch AI run | ||
| 216 | 27.5% | Jun 2024 | Epoch AI run | ||
| 217 | GPT-3.5 Turbo Jan 2024 | 27.2% | Jan 2024 | Epoch AI run | |
| 218 | 27.1% | Oct 2024 | Epoch AI run | ||
| 219 | 27% | Jul 2024 | Epoch AI run | ||
| 220 | 26.3% | Jul 2023 | Epoch AI run | ||
| 221 | 26.1% | Apr 2024 | Epoch AI run | ||
| 222 | 25.3% | Oct 2024 | Epoch AI run | ||
| 223 | 24.6% | Nov 2023 | Epoch AI run | ||
| 224 | granite-4.0-1b | 24% | Oct 2025 | Epoch AI run | |
| 225 | 23.9% | Sep 2024 | Epoch AI run | ||
| 226 | 23.2% | Mar 2025 | Epoch AI run | ||
| 227 | 19.9% | Mar 2025 | Epoch AI run | ||
| 228 | 15.2% | May 2024 | Epoch AI run | ||
| 229 | 14.7% | Nov 2023 | Epoch AI run | ||
| 230 | granite-4.0-350m | 11.2% | Oct 2025 | Epoch AI run | |
| 231 | deepseek-r1-0528-qwen3-8b | 9.28% | May 2025 | Epoch AI run |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6.2 Astra | 97% | Nov 2026 | Lab model card | |
| 2 | 97% | Oct 2026 | Official leaderboard | ||
| 3 | 96.8% | xhigh | Nov 2026 | Epoch AI run | |
| 4 | 96.4% | high | Oct 2026 | Epoch AI run | |
| 5 | Claude Fable 5.2 | 96% | with fallback, max | Nov 2026 | Epoch AI run |
| 6 | GPT-6 Astra | 96% | Sep 2026 | Lab model card | |
| 7 | GPT-6.2 Sol | 96% | Nov 2026 | Lab model card | |
| 8 | Claude Sonnet 5.5 | 95.6% | max | Sep 2026 | Epoch AI run |
| 9 | 95.4% | high | Sep 2026 | Epoch AI run | |
| 10 | GPT-6.1 Sol | 95.4% | max | Sep 2026 | Epoch AI run |
| 11 | Grok 4.6 | 95% | Aug 2026 | Official leaderboard | |
| 12 | 94.8% | high | Aug 2026 | Epoch AI run | |
| 13 | GPT-5.4 Pro | 94.6% | xhigh | Mar 2026 | Epoch AI run |
| 14 | GPT-5.6 Sol | 94.6% | Jul 2026 | Lab model card | |
| 15 | Claude Mythos Preview | 94.5% | Apr 2026 | Lab model card | |
| 16 | 94.4% | high | Feb 2026 | Epoch AI run | |
| 17 | GPT-6 Sol | 94.3% | max | Sep 2026 | Epoch AI run |
| 18 | GPT-6.1 Terra | 94.3% | max | Oct 2026 | Epoch AI run |
| 19 | Claude Opus 4.7 | 94.2% | Apr 2026 | Lab model card | |
| 20 | Kimi K3.1 Pro | 94.2% | Oct 2026 | Lab model card | |
| 21 | 94.1% | high | Jul 2026 | Epoch AI run | |
| 22 | Claude Opus 5.5 | 94.1% | with fallback, max | Sep 2026 | Epoch AI run |
| 23 | GPT-5.5 | 94% | xhigh | Apr 2026 | Epoch AI run |
| 24 | GPT-5.5 Pro | 93.9% | xhigh | Apr 2026 | Epoch AI run |
| 25 | Claude Fable 5 | 93.9% | with fallback, max | Jun 2026 | Epoch AI run |
| 26 | Claude Opus 5 | 93.9% | max | Jul 2026 | Epoch AI run |
| 27 | Claude Opus 4.8 | 93.6% | May 2026 | Lab model card | |
| 28 | Kimi K3 | 93.5% | Jul 2026 | Lab model card | |
| 29 | 93.5% | Sep 2026 | Lab model card | ||
| 30 | Grok 4.5 | 93.4% | high | Jul 2026 | Epoch AI run |
| 31 | GPT-5.6 Terra | 93.3% | max | Jul 2026 | Epoch AI run |
| 32 | GPT-5.4 | 93.3% | xhigh | Mar 2026 | Epoch AI run |
| 33 | GPT-5.2 Pro | 93.2% | Dec 2025 | Lab model card | |
| 34 | 93.2% | max | Oct 2026 | Epoch AI run | |
| 35 | 93.1% | high | Nov 2026 | Epoch AI run | |
| 36 | 93% | Nov 2026 | Lab model card | ||
| 37 | 92.8% | high | May 2026 | Epoch AI run | |
| 38 | Grok 4.7 | 92.7% | xhigh | Sep 2026 | Epoch AI run |
| 39 | 92.7% | xhigh | Aug 2026 | Epoch AI run | |
| 40 | 92.6% | Nov 2025 | Epoch AI run | ||
| 41 | GPT-5.2 | 92.4% | Dec 2025 | Lab model card | |
| 42 | 92.4% | May 2026 | Lab model card | ||
| 43 | GPT-5.6 Luna | 92.3% | Jul 2026 | Lab model card | |
| 44 | 92.3% | Aug 2026 | Lab model card | ||
| 45 | 92.3% | xhigh | Sep 2026 | Epoch AI run | |
| 46 | GLM-5.2 | 91.9% | max | Jun 2026 | Epoch AI run |
| 47 | 91.7% | Aug 2026 | Lab model card | ||
| 48 | 91.7% | max | Aug 2026 | Epoch AI run | |
| 49 | Claude Opus 4.6 | 91.3% | Feb 2026 | Lab model card | |
| 50 | 91% | max | Jul 2026 | Epoch AI run | |
| 51 | 91% | May 2026 | Lab model card | ||
| 52 | 91% | Sep 2026 | Lab model card | ||
| 53 | 90.9% | high | Apr 2026 | Epoch AI run | |
| 54 | 90.9% | Jun 2026 | Epoch AI run | ||
| 55 | GLM-5.3 | 90.9% | max | Aug 2026 | Epoch AI run |
| 56 | 90.9% | Sep 2026 | Lab model card | ||
| 57 | Kimi K2.6 | 90.8% | Apr 2026 | Epoch AI run | |
| 58 | Claude Sonnet 5 | 90.5% | xhigh | Jun 2026 | Epoch AI run |
| 59 | Beam 1.0 | 90.5% | Oct 2026 | Lab model card | |
| 60 | GPT-6 Luna | 90.5% | max | Sep 2026 | Epoch AI run |
| 61 | 90.4% | Dec 2025 | Lab model card | ||
| 62 | 90.4% | Apr 2026 | Lab model card | ||
| 63 | 90.4% | Jul 2026 | Lab model card | ||
| 64 | 90.3% | Jun 2026 | Lab model card | ||
| 65 | GLM-5.3-Flash | 90.2% | max | Aug 2026 | Epoch AI run |
| 66 | 90.2% | high | Oct 2026 | Epoch AI run | |
| 67 | Claude Sonnet 4.6 | 89.9% | Feb 2026 | Lab model card | |
| 68 | GLM-5.1 | 89.9% | Apr 2026 | Epoch AI run | |
| 69 | 89.8% | Apr 2026 | Epoch AI run | ||
| 70 | 89.8% | May 2026 | Lab model card | ||
| 71 | Claude Haiku 5.5 | 89.7% | max | Oct 2026 | Epoch AI run |
| 72 | 89.6% | high | Jun 2026 | Epoch AI run | |
| 73 | 89.5% | Oct 2026 | Lab model card | ||
| 74 | 89.5% | Jul 2026 | Lab model card | ||
| 75 | GPT-5 | 89.4% | Aug 2025 | Lab model card | |
| 76 | Grok 4.20 | 89.3% | Feb 2026 | Epoch AI run | |
| 77 | 89.3% | high | Jun 2026 | Epoch AI run | |
| 78 | Mercury 3 | 89.2% | Oct 2026 | Lab model card | |
| 79 | 89.2% | Aug 2026 | Lab model card | ||
| 80 | 89% | Aug 2026 | Lab model card | ||
| 81 | 89% | Oct 2026 | Lab model card | ||
| 82 | 88.9% | Feb 2026 | Lab model card | ||
| 83 | 88.9% | Jul 2026 | Lab model card | ||
| 84 | Grok 4.3 Beta | 88.8% | high | Apr 2026 | Epoch AI run |
| 85 | Grok 4 Heavy | 88.4% | Jul 2025 | Lab model card | |
| 86 | 88.4% | Feb 2026 | Lab model card | ||
| 87 | 88.4% | Apr 2026 | Lab model card | ||
| 88 | 88.3% | May 2026 | Lab model card | ||
| 89 | 88.3% | xhigh | Jul 2026 | Epoch AI run | |
| 90 | 88.2% | high | Feb 2026 | Epoch AI run | |
| 91 | 88.1% | high | Apr 2026 | Epoch AI run | |
| 92 | GPT-5.1 | 88.1% | Nov 2025 | Lab model card | |
| 93 | 88.1% | Apr 2026 | Lab model card | ||
| 94 | 88.1% | high | Oct 2026 | Epoch AI run | |
| 95 | GPT-5.4 Mini | 88% | Mar 2026 | Lab model card | |
| 96 | Kimi K2.7 Code | 87.9% | Jun 2026 | Epoch AI run | |
| 97 | GLM-5 | 87.8% | Feb 2026 | Epoch AI run | |
| 98 | 87.8% | Apr 2026 | Lab model card | ||
| 99 | Kimi K2.5 | 87.6% | Fireworks | Jan 2026 | Epoch AI run |
| 100 | Grok 4 | 87.5% | Jul 2025 | Lab model card | |
| 101 | 87.4% | Jan 2026 | Lab model card | ||
| 102 | 87.4% | Apr 2026 | Epoch AI run | ||
| 103 | 87.2% | Apr 2026 | Lab model card | ||
| 104 | Claude Opus 4.5 | 87% | Nov 2025 | Lab model card | |
| 105 | 87% | Jun 2026 | Lab model card | ||
| 106 | 86.9% | Mar 2026 | Lab model card | ||
| 107 | 86.6% | Feb 2026 | Lab model card | ||
| 108 | 86.6% | Nov 2026 | Lab model card | ||
| 109 | 86.4% | Oct 2026 | Lab model card | ||
| 110 | 86.4% | Jun 2025 | Lab model card | ||
| 111 | 86.4% | Feb 2026 | Lab model card | ||
| 112 | 86.3% | Jul 2026 | Lab model card | ||
| 113 | 86% | Apr 2026 | Lab model card | ||
| 114 | Grok 4 Fast | 85.7% | Sep 2025 | Lab model card | |
| 115 | 85.7% | Dec 2025 | Lab model card | ||
| 116 | GLM-4.7 | 85.7% | Dec 2025 | Lab model card | |
| 117 | GPT-5.5 Instant | 85.6% | May 2026 | Lab model card | |
| 118 | 85.5% | Feb 2026 | Lab model card | ||
| 119 | 85.2% | Jan 2026 | Lab model card | ||
| 120 | 85.2% | Feb 2026 | Lab model card | ||
| 121 | 84.8% | Feb 2026 | Epoch AI run | ||
| 122 | Claude 3.7 Sonnet | 84.8% | Feb 2025 | Lab model card | |
| 123 | Grok 3 | 84.6% | Feb 2025 | Lab model card | |
| 124 | 84.6% | Jun 2026 | Lab model card | ||
| 125 | Kimi K2 Thinking | 84.5% | Nov 2025 | Lab model card | |
| 126 | MiMo-V2-Flash | 84.3% | Jan 2026 | Lab model card | |
| 127 | 84.3% | Apr 2026 | Lab model card | ||
| 128 | 84.2% | Feb 2026 | Lab model card | ||
| 129 | 84.2% | Jun 2026 | Lab model card | ||
| 130 | Grok-3 mini | 84% | Feb 2025 | Lab model card | |
| 131 | 84% | Mar 2025 | Lab model card | ||
| 132 | o3-pro | 84% | Jun 2025 | Lab model card | |
| 133 | 83.9% | Mar 2026 | Lab model card | ||
| 134 | Claude Sonnet 4 | 83.8% | May 2025 | Lab model card | |
| 135 | 83.8% | Dec 2025 | Lab model card | ||
| 136 | 83.5% | Feb 2026 | Lab model card | ||
| 137 | 83.5% | Aug 2026 | Lab model card | ||
| 138 | 83.4% | Thinking | Dec 2025 | Epoch AI run | |
| 139 | Claude Sonnet 4.5 | 83.4% | Sep 2025 | Lab model card | |
| 140 | 83.3% | 2026-04-16 | Apr 2026 | Epoch AI run | |
| 141 | 83.3% | high | Jul 2026 | Epoch AI run | |
| 142 | o3 | 83.3% | Apr 2025 | Lab model card | |
| 143 | Claude Opus 4 | 83.3% | May 2025 | Lab model card | |
| 144 | 83.3% | Sep 2025 | Lab model card | ||
| 145 | 83.2% | Apr 2026 | Official leaderboard | ||
| 146 | 83% | Dec 2025 | Lab model card | ||
| 147 | 82.8% | Jun 2025 | Lab model card | ||
| 148 | GPT-5.4 Nano | 82.8% | Mar 2026 | Lab model card | |
| 149 | 82.3% | Feb 2026 | Epoch AI run | ||
| 150 | 82.3% | Jul 2026 | Epoch AI run | ||
| 151 | GPT-5 mini | 82.3% | Aug 2025 | Lab model card | |
| 152 | 82.2% | Jul 2026 | Lab model card | ||
| 153 | 81.7% | Mar 2026 | Lab model card | ||
| 154 | 81.5% | Jun 2025 | Lab model card | ||
| 155 | 81.5% | Sep 2025 | Lab model card | ||
| 156 | o4-mini | 81.4% | Apr 2025 | Lab model card | |
| 157 | 81.4% | Dec 2025 | Lab model card | ||
| 158 | 81.1% | Jul 2025 | Lab model card | ||
| 159 | 81% | May 2025 | Lab model card | ||
| 160 | GLM-4.6 | 81% | Sep 2025 | Lab model card | |
| 161 | Claude Opus 4.1 | 80.9% | Aug 2025 | Lab model card | |
| 162 | 80.8% | Sep 2025 | Lab model card | ||
| 163 | 80.7% | Sep 2025 | Lab model card | ||
| 164 | 80.5% | Apr 2026 | Lab model card | ||
| 165 | gpt-oss-120b | 80.1% | Aug 2025 | Lab model card | |
| 166 | 80.1% | Aug 2025 | Lab model card | ||
| 167 | 79.9% | Sep 2025 | Lab model card | ||
| 168 | o3-mini | 79.7% | Jan 2025 | Lab model card | |
| 169 | 79.6% | Dec 2025 | Lab model card | ||
| 170 | 79.2% | Mar 2026 | Lab model card | ||
| 171 | GLM-4.5 | 79.1% | Jul 2025 | Lab model card | |
| 172 | 79.1% | Dec 2025 | Lab model card | ||
| 173 | Mercury 2.5 | 79% | Sep 2026 | Lab model card | |
| 174 | 78.8% | Jun 2026 | Lab model card | ||
| 175 | Sarvam 105B | 78.7% | Mar 2026 | Lab model card | |
| 176 | 78.6% | Oct 2025 | Lab model card | ||
| 177 | o1-preview | 78.3% | Sep 2024 | Lab model card | |
| 178 | 78.3% | Apr 2025 | Lab model card | ||
| 179 | 78% | Oct 2025 | Lab model card | ||
| 180 | 77.7% | Nov 2025 | Lab model card | ||
| 181 | 77.6% | thinking | Jun 2025 | Epoch AI run | |
| 182 | 77.3% | Apr 2025 | Lab model card | ||
| 183 | 77.3% | Jul 2025 | Lab model card | ||
| 184 | 77.2% | Sep 2025 | Lab model card | ||
| 185 | 77.1% | Nov 2025 | Lab model card | ||
| 186 | 77.1% | Oct 2026 | Lab model card | ||
| 187 | o1 | 76.8% | high | Dec 2024 | Epoch AI run |
| 188 | 76.5% | Oct 2026 | Lab model card | ||
| 189 | 76.3% | Apr 2026 | Lab model card | ||
| 190 | 76.3% | Sep 2025 | Lab model card | ||
| 191 | 76.2% | Mar 2026 | Lab model card | ||
| 192 | 76.2% | Jun 2026 | Lab model card | ||
| 193 | 76.1% | Mar 2026 | Lab model card | ||
| 194 | 76% | Jul 2025 | Lab model card | ||
| 195 | 76% | Apr 2025 | Lab model card | ||
| 196 | 75.4% | Jul 2025 | Lab model card | ||
| 197 | glm-4.7-flash | 75.2% | Jan 2026 | Lab model card | |
| 198 | Kimi K2 Jul 2025 | 75.1% | Jul 2025 | Lab model card | |
| 199 | GLM-4.5-Air | 75% | Aug 2025 | Lab model card | |
| 200 | 74.5% | Feb 2026 | Lab model card | ||
| 201 | LFM2.5-24B-A2B | 74.4% | Oct 2026 | Lab model card | |
| 202 | 74.4% | Oct 2025 | Lab model card | ||
| 203 | INTELLECT-3 | 74.4% | Nov 2025 | Lab model card | |
| 204 | 74.2% | Jan 2025 | Lab model card | ||
| 205 | Mercury 2 | 74% | Feb 2026 | Lab model card | |
| 206 | 73.7% | Apr 2025 | Lab model card | ||
| 207 | 73.2% | Aug 2025 | Lab model card | ||
| 208 | 73.2% | Aug 2025 | Lab model card | ||
| 209 | 73.2% | Jun 2026 | Lab model card | ||
| 210 | 73% | Dec 2025 | Lab model card | ||
| 211 | 73% | Jul 2025 | Lab model card | ||
| 212 | Claude Haiku 4.5 | 73% | Oct 2025 | Lab model card | |
| 213 | 72.6% | Sep 2025 | Epoch AI run | ||
| 214 | Beam 1.0 Base | 72.4% | Oct 2026 | Lab model card | |
| 215 | 72.2% | Apr 2026 | Lab model card | ||
| 216 | 72% | Jul 2025 | Lab model card | ||
| 217 | 71.9% | Mar 2026 | Lab model card | ||
| 218 | 71.7% | Jan 2025 | Epoch AI run | ||
| 219 | 71.7% | Sep 2025 | Lab model card | ||
| 220 | 71.5% | Aug 2025 | Epoch AI run | ||
| 221 | gpt-oss-20b | 71.5% | Aug 2025 | Lab model card | |
| 222 | 71.5% | Jul 2025 | Lab model card | ||
| 223 | GPT-4.5 | 71.4% | Feb 2025 | Lab model card | |
| 224 | 71.2% | Jun 2025 | Lab model card | ||
| 225 | GPT-5 nano | 71.2% | Aug 2025 | Lab model card | |
| 226 | 71.2% | Dec 2025 | Lab model card | ||
| 227 | 71.2% | Mar 2026 | Lab model card | ||
| 228 | 71.1% | Apr 2025 | Lab model card | ||
| 229 | 70.8% | Jan 2026 | Lab model card | ||
| 230 | 70.8% | Jun 2025 | Lab model card | ||
| 231 | Hermes 4 405B | 70.6% | Aug 2025 | Lab model card | |
| 232 | 70.1% | Jul 2025 | Epoch AI run | ||
| 233 | 70.1% | Sep 2025 | Lab model card | ||
| 234 | 70% | Jun 2025 | Lab model card | ||
| 235 | 69.8% | Apr 2025 | Lab model card | ||
| 236 | 69.6% | Sep 2025 | Lab model card | ||
| 237 | 69.3% | Mar 2025 | Lab model card | ||
| 238 | 69.3% | Apr 2025 | Lab model card | ||
| 239 | 68.4% | May 2025 | Lab model card | ||
| 240 | 68.2% | Jun 2025 | Lab model card | ||
| 241 | 68.1% | Jan 2026 | Lab model card | ||
| 242 | 67.6% | Mar 2025 | Epoch AI run | ||
| 243 | 67.1% | Apr 2025 | Lab model card | ||
| 244 | GPT-4.1 | 66.9% | Apr 2025 | Epoch AI run | |
| 245 | INTELLECT-2 | 66.8% | May 2025 | Lab model card | |
| 246 | 66.7% | May 2025 | Lab model card | ||
| 247 | 66.7% | Jun 2025 | Lab model card | ||
| 248 | 66.7% | May 2025 | Epoch AI run | ||
| 249 | Sarvam 30B | 66.5% | Mar 2026 | Lab model card | |
| 250 | Granite 4.2 30B | 66.4% | Aug 2026 | Lab model card | |
| 251 | 66.2% | Oct 2026 | Lab model card | ||
| 252 | 66.1% | Mar 2025 | Lab model card | ||
| 253 | GLM-Z1-32B-0414 | 66.1% | Apr 2025 | Lab model card | |
| 254 | GPT-4.1 mini | 65.8% | Apr 2025 | Epoch AI run | |
| 255 | 65.8% | May 2025 | Lab model card | ||
| 256 | 65.8% | Jul 2025 | Lab model card | ||
| 257 | 65.7% | Feb 2025 | Epoch AI run | ||
| 258 | Hermes 4.3 36B | 65.5% | Dec 2025 | Lab model card | |
| 259 | 65.4% | Apr 2025 | Epoch AI run | ||
| 260 | 65.3% | Mar 2025 | Epoch AI run | ||
| 261 | 65.2% | Jan 2025 | Lab model card | ||
| 262 | Claude 3.5 Sonnet (October 2024) | 65% | Oct 2024 | Lab model card | |
| 263 | 64.1% | Jan 2025 | Epoch AI run | ||
| 264 | 64.1% | Feb 2025 | Epoch AI run | ||
| 265 | 64.1% | Nov 2025 | Lab model card | ||
| 266 | 64% | May 2025 | Lab model card | ||
| 267 | nvidia-nemotron-nano-9b-v2 | 64% | Aug 2025 | Lab model card | |
| 268 | LFM3-1.7B | 63.4% | Oct 2026 | Lab model card | |
| 269 | 63.3% | Feb 2026 | Lab model card | ||
| 270 | o1-mini | 62.4% | high | Sep 2024 | Epoch AI run |
| 271 | 62.1% | Dec 2024 | Lab model card | ||
| 272 | 62% | May 2025 | Lab model card | ||
| 273 | 61.3% | Jan 2026 | Lab model card | ||
| 274 | Reka Flash 3 | 61.1% | Mar 2025 | Lab model card | |
| 275 | 60.6% | Apr 2025 | Lab model card | ||
| 276 | 59.5% | May 2025 | Epoch AI run | ||
| 277 | Claude 3.5 Sonnet | 59.4% | Jun 2024 | Lab model card | |
| 278 | 59.1% | Dec 2024 | Lab model card | ||
| 279 | 59.1% | Jan 2025 | Lab model card | ||
| 280 | 58.6% | Apr 2026 | Lab model card | ||
| 281 | 58.5% | Dec 2025 | Lab model card | ||
| 282 | 58.1% | Nov 2025 | Lab model card | ||
| 283 | 57.5% | Feb 2025 | Lab model card | ||
| 284 | 57.5% | Dec 2025 | Lab model card | ||
| 285 | 57.2% | Sep 2024 | Epoch AI run | ||
| 286 | 57.2% | Apr 2025 | Lab model card | ||
| 287 | 57.1% | Apr 2025 | Lab model card | ||
| 288 | 56.1% | Jan 2025 | Epoch AI run | ||
| 289 | 56.1% | Dec 2024 | Lab model card | ||
| 290 | Grok-2 Aug 2024 | 56% | Aug 2024 | Lab model card | |
| 291 | 55.9% | May 2025 | Lab model card | ||
| 292 | 55.6% | Jul 2025 | Epoch AI run | ||
| 293 | 54.4% | Jan 2025 | Lab model card | ||
| 294 | MiMo-7B-RL | 54.4% | Apr 2025 | Lab model card | |
| 295 | Grok-2 Dec 2024 | 53.8% | Dec 2024 | Epoch AI run | |
| 296 | GPT-4o May 2024 | 53.6% | May 2024 | Lab model card | |
| 297 | 53.2% | Mar 2026 | Lab model card | ||
| 298 | 52% | Apr 2025 | Lab model card | ||
| 299 | 51.6% | Mar 2026 | Lab model card | ||
| 300 | 51.5% | Feb 2025 | Lab model card | ||
| 301 | 51.3% | Nov 2024 | Epoch AI run | ||
| 302 | 51.1% | Jul 2024 | Lab model card | ||
| 303 | Mercury | 51% | Jun 2025 | Lab model card | |
| 304 | 50.8% | Apr 2025 | Lab model card | ||
| 305 | 50.5% | Dec 2024 | Lab model card | ||
| 306 | Claude 3 Opus | 50.4% | Mar 2024 | Lab model card | |
| 307 | GPT-4.1 nano | 50.3% | Apr 2025 | Lab model card | |
| 308 | 50% | Jun 2026 | Lab model card | ||
| 309 | 49.7% | May 2025 | Lab model card | ||
| 310 | 49.5% | Dec 2024 | Lab model card | ||
| 311 | GPT-4 Turbo Apr 2024 | 49.3% | Apr 2024 | Lab model card | |
| 312 | GPT-4o Aug 2024 | 49.2% | Aug 2024 | Epoch AI run | |
| 313 | 49.1% | Sep 2024 | Epoch AI run | ||
| 314 | 49% | Jun 2025 | Epoch AI run | ||
| 315 | 49% | Jul 2024 | Epoch AI run | ||
| 316 | 49% | Feb 2025 | Lab model card | ||
| 317 | 48.1% | Jan 2025 | Epoch AI run | ||
| 318 | Sarvam-M | 48% | May 2025 | Lab model card | |
| 319 | GPT-4o Nov 2024 | 47.9% | Nov 2024 | Epoch AI run | |
| 320 | 47.7% | Mar 2025 | Epoch AI run | ||
| 321 | 47.5% | Mar 2025 | Epoch AI run | ||
| 322 | 47.4% | Dec 2024 | Epoch AI run | ||
| 323 | 47.3% | Sep 2024 | Epoch AI run | ||
| 324 | 47.3% | Jan 2025 | Epoch AI run | ||
| 325 | 46.7% | Jul 2024 | Lab model card | ||
| 326 | 46.3% | Nov 2024 | Epoch AI run | ||
| 327 | 46.2% | May 2024 | Lab model card | ||
| 328 | qwen3-4b-instruct-2507 | 45.8% | Aug 2025 | Epoch AI run | |
| 329 | Granite 4.1 30B | 45.8% | Apr 2026 | Lab model card | |
| 330 | 45.5% | Dec 2024 | Lab model card | ||
| 331 | glm-4.7-flash_none | 45.1% | none | Epoch AI run | |
| 332 | 45.1% | Jul 2025 | Lab model card | ||
| 333 | 43.9% | Dec 2025 | Lab model card | ||
| 334 | 43.4% | Apr 2024 | Epoch AI run | ||
| 335 | 43.4% | Apr 2026 | Lab model card | ||
| 336 | 42.4% | Jun 2024 | Lab model card | ||
| 337 | GPT-4 Turbo Nov 2023 | 42.4% | Nov 2023 | Epoch AI run | |
| 338 | 42.3% | Dec 2024 | Lab model card | ||
| 339 | Kimi-VL-A3B-Thinking-2506 | 42.3% | Jun 2025 | Lab model card | |
| 340 | Claude 3.5 Haiku | 41.6% | Oct 2024 | Lab model card | |
| 341 | 41.5% | Feb 2024 | Lab model card | ||
| 342 | 41% | Sep 2024 | Epoch AI run | ||
| 343 | 40.9% | Mar 2025 | Lab model card | ||
| 344 | Granite 4.0 H Small | 40.6% | Oct 2025 | Lab model card | |
| 345 | 40.6% | Sep 2025 | Lab model card | ||
| 346 | Claude 3 Sonnet | 40.6% | Feb 2024 | Epoch AI run | |
| 347 | 40.6% | Apr 2024 | Epoch AI run | ||
| 348 | 40.4% | May 2024 | Epoch AI run | ||
| 349 | GPT-4o mini | 40.2% | Jul 2024 | Lab model card | |
| 350 | 40.1% | May 2025 | Lab model card | ||
| 351 | 39.9% | Jan 2025 | Lab model card | ||
| 352 | LFM2.5-1.2B-Instruct | 38.9% | Jan 2026 | Lab model card | |
| 353 | 38.8% | Feb 2024 | Epoch AI run | ||
| 354 | LFM2.5-1.2B-Thinking | 37.9% | Jan 2026 | Lab model card | |
| 355 | Hermes 2 Theta Llama-3 70B | 37.5% | Jun 2024 | Epoch AI run | |
| 356 | 37.4% | Jan 2025 | Lab model card | ||
| 357 | 36.8% | Aug 2024 | Lab model card | ||
| 358 | 36.5% | Jun 2024 | Epoch AI run | ||
| 359 | 36.4% | Dec 2024 | Lab model card | ||
| 360 | Claude 3 Haiku | 36.3% | Mar 2024 | Epoch AI run | |
| 361 | 35.9% | Dec 2024 | Lab model card | ||
| 362 | GPT-4 Mar 2023 | 35.7% | Mar 2023 | Epoch AI run | |
| 363 | GLM-4 | 35.7% | Jun 2024 | Lab model card | |
| 364 | Claude 2 | 34.7% | Jul 2023 | Epoch AI run | |
| 365 | 34.3% | Aug 2024 | Lab model card | ||
| 366 | 34.2% | Apr 2024 | Lab model card | ||
| 367 | 34.1% | Apr 2024 | Epoch AI run | ||
| 368 | 34% | Dec 2023 | Epoch AI run | ||
| 369 | Eurus-2-7B-PRIME | 33.9% | Dec 2024 | Epoch AI run | |
| 370 | 33.8% | Jan 2025 | Lab model card | ||
| 371 | 33.3% | Oct 2026 | Lab model card | ||
| 372 | Claude 2.1 | 33% | Nov 2023 | Epoch AI run | |
| 373 | 33% | Oct 2024 | Epoch AI run | ||
| 374 | 32.9% | Mar 2024 | Epoch AI run | ||
| 375 | 32.8% | Aug 2024 | Lab model card | ||
| 376 | 32.8% | Nov 2024 | Lab model card | ||
| 377 | 32% | May 2024 | Epoch AI run | ||
| 378 | LFM2-1.2B | 31.5% | Jul 2025 | Lab model card | |
| 379 | 30.8% | Mar 2025 | Lab model card | ||
| 380 | 30.8% | Mar 2025 | Lab model card | ||
| 381 | 30.7% | Apr 2024 | Epoch AI run | ||
| 382 | GPT-4 Jun 2023 | 30.6% | Jun 2023 | Epoch AI run | |
| 383 | 30.6% | Dec 2023 | Epoch AI run | ||
| 384 | 30.4% | Feb 2025 | Lab model card | ||
| 385 | granite-4.0-1b | 29.9% | Oct 2025 | Lab model card | |
| 386 | 29.9% | Jul 2024 | Epoch AI run | ||
| 387 | LFM2-8B-A1B | 29.3% | Oct 2025 | Lab model card | |
| 388 | 28.8% | Feb 2024 | Epoch AI run | ||
| 389 | granite-4.0-micro | 28.3% | Oct 2025 | Epoch AI run | |
| 390 | GPT-3.5 Turbo Nov 2023 | 28% | Nov 2023 | Epoch AI run | |
| 391 | 27.8% | Jul 2025 | Lab model card | ||
| 392 | 27.6% | Apr 2024 | Epoch AI run | ||
| 393 | 27.5% | Jun 2024 | Epoch AI run | ||
| 394 | 27.2% | Aug 2024 | Lab model card | ||
| 395 | GPT-3.5 Turbo Jan 2024 | 27.2% | Jan 2024 | Epoch AI run | |
| 396 | 27.1% | Oct 2024 | Epoch AI run | ||
| 397 | ChatGLM3-6B | 26.8% | Jun 2024 | Lab model card | |
| 398 | 26.3% | Jul 2023 | Epoch AI run | ||
| 399 | 25.3% | Oct 2024 | Epoch AI run | ||
| 400 | 24.6% | Nov 2023 | Epoch AI run | ||
| 401 | 23.9% | Sep 2024 | Epoch AI run | ||
| 402 | 23.7% | Jun 2025 | Lab model card | ||
| 403 | 19.9% | Mar 2025 | Epoch AI run | ||
| 404 | 15.2% | May 2024 | Epoch AI run | ||
| 405 | 14.7% | Nov 2023 | Epoch AI run | ||
| 406 | 11.9% | Mar 2026 | Lab model card | ||
| 407 | granite-4.0-350m | 11.2% | Oct 2025 | Epoch AI run | |
| 408 | deepseek-r1-0528-qwen3-8b | 9.28% | May 2025 | Epoch AI run |
Epoch AI has run 231 models. Labs report scores for 306 models in their cards. Switch the scores above to rank them together.
3 models have held number one on the official leaderboard since its release in November 2023 21 models have held number one across all independent runs since its release in November 2023 22 models have held number one with claims included since its release in November 2023
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard, Epoch AI’s runs and what labs claim in their cards.
Sources: Official GPQA Diamond leaderboard, Epoch AI runs (CC BY 4.0) and labs’ model cards
Google DeepMind’s own number for Gemini 2.5 Pro passed 85% in June 2025
That was 19 months after its release. By the same age, 32% of the benchmarks from 2023 had passed 85%. All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official GPQA Diamond leaderboard, Epoch AI runs (CC BY 4.0) and labs’ model cards
20 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Nov 5, 2026 | 93 | GPQA Diamond, 93.0% | |
| Oct 27, 2026 | LFM3-1.7B | 63.4 | GPQA Diamond |
| Oct 24, 2026 | Kimi K3.1 Pro | 94.2 | GPQA-Diamond |
| Oct 22, 2026 | 77.1 | GPQA Diamond | |
| Oct 19, 2026 | Mercury 3 | 89.2 | GPQA Diamond |
| Oct 15, 2026 | 89 | GPQA Diamond | |
| Oct 15, 2026 | 86.4 | GPQA Diamond | |
| Oct 15, 2026 | 89.5 | GPQA-Diamond | |
| Oct 13, 2026 | LFM2.5-24B-A2B | 74.4 | GPQA Diamond |
| Oct 5, 2026 | Beam 1.0 | 90.5 | GPQA Diamond |
| Sep 20, 2026 | 93.5 | GPQA Diamond | |
| Sep 18, 2026 | 91 | GPQA Diamond | |
| Sep 10, 2026 | 90.9 | GPQA Diamond (Pass@1) | |
| Sep 8, 2026 | Mercury 2.5 | 79 | GPQA Diamond |
| Sep 3, 2026 | GPT-6 Astra | 96 | GPQA Diamond |
| Aug 28, 2026 | 92.3 | GPQA Diamond | |
| Aug 26, 2026 | 91.7 | Scientific reasoning GPQA Diamond | |
| Aug 25, 2026 | Granite 4.2 30B | 66.4 | GPQA |
| Aug 14, 2026 | 89.2 | Scientific reasoning GPQA Diamond | |
| Aug 11, 2026 | 75.4 | GPQA Diamond (no tools) | |
| Aug 11, 2026 | 89 | GPQA Diamond (Artificial Analysis score quoted by Upstage) | |
| Aug 10, 2026 | 83.5 | GPQA Diamond (AA) | |
| Aug 3, 2026 | 92.6 | GPQA Diamond | |
| Jul 31, 2026 | 82.2 | GPQA-Diamond | |
| Jul 30, 2026 | 89.5 | GPQA Diamond | |
| Jul 22, 2026 | 86.3 | GPQA-Diamond | |
| Jul 16, 2026 | Kimi K3 | 93.5 | GPQA Diamond |
| Jul 15, 2026 | 87.2 | GPQA Diamond (Artificial Analysis) | |
| Jul 9, 2026 | GPT-5.6 Luna | 92.3 | GPQA Diamond |
| Jul 9, 2026 | GPT-5.6 Terra | 92.9 | GPQA Diamond |
| Jul 9, 2026 | GPT-5.6 Sol | 94.6 | GPQA Diamond |
| Jul 5, 2026 | 88.9 | GPQA-diamond | |
| Jul 2, 2026 | 90.4 | GPQA Diamond | |
| Jun 16, 2026 | GLM-5.2 | 91.2 | GPQA-Diamond |
| Jun 10, 2026 | 73.2 | GPQA Diamond | |
| Jun 4, 2026 | 87 | GPQA (no tools) | |
| Jun 3, 2026 | 78.8 | GPQA Diamond | |
| Jun 2, 2026 | 84.6 | GPQA Diamond | |
| Jun 2, 2026 | 84.2 | GPQA Diamond | |
| Jun 1, 2026 | 90.3 | GPQA Diamond | |
| May 28, 2026 | Claude Opus 4.8 | 93.6 | GPQA Diamond |
| May 20, 2026 | 92.4 | GPQA Diamond | |
| May 14, 2026 | 88.3 | GPQA Diamond (xhigh) | |
| May 9, 2026 | 91 | GPQA (pass@1) | |
| May 5, 2026 | GPT-5.5 Instant | 85.6 | GPQA (PhD-Level Science) |
| Apr 29, 2026 | Granite 4.1 30B | 45.8 | GPQA (0-shot, CoT; subset not stated) |
| Apr 28, 2026 | 88.4 | GPQA Diamond | |
| Apr 24, 2026 | 88.1 | GPQA Diamond (Pass@1), V4-Flash Max | |
| Apr 24, 2026 | 90.1 | GPQA Diamond (Pass@1) | |
| Apr 23, 2026 | 87.2 | GPQA-Diamond | |
| Apr 23, 2026 | GPT-5.5 | 93.6 | GPQA Diamond |
| Apr 22, 2026 | 87.8 | GPQA Diamond | |
| Apr 20, 2026 | Kimi K2.6 | 90.5 | GPQA-Diamond |
| Apr 16, 2026 | Claude Opus 4.7 | 94.2 | Graduate-level reasoning GPQA Diamond |
| Apr 15, 2026 | 86 | GPQA | |
| Apr 9, 2026 | 80.5 | GPQA-Diamond | |
| Apr 8, 2026 | 89.5 | GPQA Diamond (PhD Level Reasoning) | |
| Apr 7, 2026 | GLM-5.1 | 86.2 | GPQA-Diamond |
| Apr 7, 2026 | Claude Mythos Preview | 94.5 | GPQA Diamond |
| Apr 2, 2026 | 58.6 | GPQA Diamond | |
| Apr 2, 2026 | 43.4 | GPQA Diamond | |
| Apr 2, 2026 | 82.3 | GPQA Diamond | |
| Apr 2, 2026 | 90.4 | GPQA | |
| Apr 2, 2026 | 84.3 | GPQA Diamond | |
| Apr 1, 2026 | 76.3 | GPQA-Diamond | |
| Mar 30, 2026 | 83.9 | GPQA | |
| Mar 24, 2026 | 71.9 | GPQA-Diamond | |
| Mar 19, 2026 | 76.1 | GPQA-Diamond | |
| Mar 17, 2026 | GPT-5.4 Nano | 82.8 | GPQA Diamond |
| Mar 17, 2026 | GPT-5.4 Mini | 88 | GPQA Diamond |
| Mar 16, 2026 | 53.2 | GPQA (reasoning on) | |
| Mar 16, 2026 | 71.2 | GPQA Diamond (reasoning; instruct = 59.1) | |
| Mar 11, 2026 | 79.2 | GPQA (no tools) | |
| Mar 6, 2026 | Sarvam 30B | 66.5 | GPQA Diamond |
| Mar 6, 2026 | Sarvam 105B | 78.7 | GPQA Diamond |
| Mar 5, 2026 | GPT-5.4 Pro | 94.4 | GPQA Diamond |
| Mar 5, 2026 | GPT-5.4 | 92.8 | GPQA Diamond (GPT-5.4) |
| Mar 3, 2026 | 86.9 | GPQA Diamond (No tools) | |
| Mar 2, 2026 | 11.9 | GPQA | |
| Mar 2, 2026 | 51.6 | GPQA | |
| Mar 2, 2026 | 76.2 | GPQA Diamond | |
| Mar 2, 2026 | 81.7 | GPQA Diamond | |
| Feb 24, 2026 | 84.2 | GPQA Diamond | |
| Feb 24, 2026 | 85.5 | GPQA Diamond | |
| Feb 24, 2026 | Mercury 2 | 74 | GPQA Diamond |
| Feb 24, 2026 | 86.6 | GPQA Diamond | |
| Feb 19, 2026 | 94.3 | GPQA Diamond (No tools) | |
| Feb 17, 2026 | Claude Sonnet 4.6 | 89.9 | Graduate-level reasoning GPQA Diamond |
| Feb 16, 2026 | 88.4 | GPQA | |
| Feb 14, 2026 | 88.9 | GPQA Diamond | |
| Feb 12, 2026 | 85.2 | GPQA-D | |
| Feb 11, 2026 | GLM-5 | 86 | GPQA-Diamond |
| Feb 6, 2026 | 86.4 | GPQA-Diamond | |
| Feb 5, 2026 | Claude Opus 4.6 | 91.3 | Graduate-level reasoning GPQA Diamond |
| Jan 27, 2026 | 63.3 | GPQA-Diamond | |
| Jan 27, 2026 | Kimi K2.5 | 87.6 | GPQA-Diamond |
| Jan 23, 2026 | 87.4 | GPQA | |
| Jan 20, 2026 | LFM2.5-1.2B-Thinking | 37.9 | GPQA Diamond (± 0.83) |
| Jan 19, 2026 | glm-4.7-flash | 75.2 | GPQA |
| Jan 14, 2026 | 85.2 | Best of GPQA-Diamond (Avg@16) 80.5 / Heavy Thinking 85.2 | |
| Jan 5, 2026 | 61.3 | GPQA-D | |
| Jan 5, 2026 | LFM2.5-1.2B-Instruct | 38.9 | GPQA |
| Jan 5, 2026 | 68.1 | GPQA-Diamond | |
| Dec 31, 2025 | 79.1 | GPQA-Diamond | |
| Dec 23, 2025 | 83 | GPQA-D | |
| Dec 22, 2025 | GLM-4.7 | 85.7 | GPQA-Diamond |
| Dec 18, 2025 | 83.8 | GPQA-Diamond | |
| Dec 17, 2025 | 90.4 | GPQA Diamond (No tools) | |
| Dec 16, 2025 | MiMo-V2-Flash | 83.7 | GPQA-Diamond |
| Dec 15, 2025 | 73 | GPQA (no tools) | |
| Dec 12, 2025 | 57.5 | GPQA | |
| Dec 11, 2025 | GPT-5.2 Pro | 93.2 | GPQA Diamond (no tools) |
| Dec 11, 2025 | GPT-5.2 | 92.4 | GPQA Diamond (no tools) Science questions |
| Dec 3, 2025 | Hermes 4.3 36B | 65.5 | GPQA Diamond |
| Dec 2, 2025 | 71.2 | GPQA Diamond 0.712 (reasoning) | |
| Dec 2, 2025 | 79.6 | GPQA Diamond acc | |
| Dec 2, 2025 | 81.4 | GPQA Diamond acc | |
| Dec 2, 2025 | 43.9 | GPQA-Diamond (5-shot, no CoT) [base] | |
| Dec 1, 2025 | 58.5 | GPQA-Diamond | |
| Dec 1, 2025 | 82.4 | GPQA Diamond (Pass@1) | |
| Nov 26, 2025 | INTELLECT-3 | 74.4 | GPQA |
| Nov 24, 2025 | Claude Opus 4.5 | 87 | Graduate-level reasoning GPQA Diamond |
| Nov 20, 2025 | 58.1 | GPQA | |
| Nov 19, 2025 | 77.7 | GPQA Diamond | |
| Nov 18, 2025 | 91.9 | GPQA Diamond (No tools) | |
| Nov 13, 2025 | GPT-5.1 | 88.1 | GPQA Diamond (no tools) |
| Nov 6, 2025 | Kimi K2 Thinking | 84.5 | GPQA, no tools |
| Oct 31, 2025 | 74.4 | GPQA-diamond(acc) | |
| Oct 28, 2025 | granite-4.0-1b | 29.9 | GPQA (0-shot, CoT; subset not stated) |
| Oct 27, 2025 | 78 | GPQA-Diamond (AA) | |
| Oct 15, 2025 | Claude Haiku 4.5 | 73 | Graduate-level reasoning GPQA Diamond |
| Oct 7, 2025 | LFM2-8B-A1B | 29.3 | GPQA |
| Oct 2, 2025 | Granite 4.0 H Small | 40.6 | GPQA (0-shot, CoT; subset not stated) |
| Sep 30, 2025 | GLM-4.6 | 81 | GPQA |
| Sep 29, 2025 | 79.9 | GPQA-Diamond | |
| Sep 29, 2025 | Claude Sonnet 4.5 | 83.4 | Graduate-level reasoning GPQA Diamond |
| Sep 25, 2025 | 71.7 | GPQA diamond, Thinking | |
| Sep 25, 2025 | 80.8 | GPQA diamond single attempt, model card | |
| Sep 24, 2025 | GPQA (Thinking Heavy w/ Python) | ||
| Sep 22, 2025 | 69.6 | GPQA | |
| Sep 22, 2025 | 80.7 | GPQA-Diamond | |
| Sep 21, 2025 | 81.5 | GPQA-Diamond(Mean@16) | |
| Sep 19, 2025 | Grok 4 Fast | 85.7 | GPQA Diamond pass@1 |
| Sep 17, 2025 | 70.1 | GPQA Diamond | |
| Sep 17, 2025 | 76.3 | GPQA Diamond | |
| Sep 11, 2025 | 77.2 | GPQA | |
| Aug 29, 2025 | 73.2 | GPQA-diamond(acc) | |
| Aug 26, 2025 | Hermes 4 405B | 70.5 | GPQA Diamond (reasoning) |
| Aug 25, 2025 | 73.2 | GPQA | |
| Aug 21, 2025 | 80.1 | GPQA-Diamond (Pass@1) | |
| Aug 20, 2025 | 71.4 | GPQA-D | |
| Aug 18, 2025 | 64 | GPQA | |
| Aug 7, 2025 | GPT-5 nano | 71.2 | GPQA diamond (no tools) |
| Aug 7, 2025 | GPT-5 mini | 82.3 | GPQA diamond (no tools) |
| Aug 7, 2025 | GPT-5 | 89.4 | GPQA Diamond, GPT-5 pro (python) |
| Aug 5, 2025 | gpt-oss-20b | 71.5 | GPQA Diamond (without tools) |
| Aug 5, 2025 | Claude Opus 4.1 | 80.9 | Graduate-level reasoning GPQA Diamond |
| Aug 5, 2025 | gpt-oss-120b | 80.1 | GPQA Diamond (without tools) |
| Jul 31, 2025 | 76 | GPQA - Diamond (thinking) | |
| Jul 31, 2025 | 73 | GPQA-Diamond | |
| Jul 29, 2025 | 27.8 | GPQA-D-Zero | |
| Jul 28, 2025 | GLM-4.5 | 79.1 | GPQA |
| Jul 25, 2025 | 72 | GPQA (pass@1, avg of 4) | |
| Jul 25, 2025 | 81.1 | GPQA | |
| Jul 24, 2025 | 65.8 | GPQA Diamond | |
| Jul 24, 2025 | 71.5 | GPQA Diamond | |
| Jul 24, 2025 | 77.3 | GPQA | |
| Jul 15, 2025 | 75.4 | GPQA-Diamond (reasoning) | |
| Jul 11, 2025 | Kimi K2 | 75.1 | GPQA-Diamond, Avg@8 |
| Jul 10, 2025 | LFM2-1.2B | 31.5 | GPQA |
| Jul 9, 2025 | 45.1 | GPQA Diamond | |
| Jul 9, 2025 | Grok 4 | 87.5 | GPQA (Science), no tools |
| Jul 9, 2025 | Grok 4 Heavy | 88.4 | GPQA (Science) [Grok 4 Heavy] |
| Jun 27, 2025 | 71.2 | GPQA-Diamond | |
| Jun 26, 2025 | 23.7 | GPQA Diamond 0-shot (IT) | |
| Jun 26, 2025 | Mercury | 51 | GPQA Diamond (% accuracy) |
| Jun 25, 2025 | 81.5 | GPQA-Diamond | |
| Jun 21, 2025 | Kimi-VL-A3B-Thinking-2506 | 42.3 | GPQA-Diamond |
| Jun 20, 2025 | 46.1 | GPQA Diamond (5-shot CoT) | |
| Jun 17, 2025 | 82.8 | GPQA diamond single attempt | |
| Jun 17, 2025 | 66.7 | GPQA diamond, Thinking | |
| Jun 17, 2025 | 86.4 | GPQA diamond, single attempt (pass@1) | |
| Jun 16, 2025 | 70 | GPQA Diamond | |
| Jun 10, 2025 | 68.2 | GPQA Diamond | |
| Jun 10, 2025 | o3-pro | 84 | PhD Science Questions (GPQA Diamond), pass@1 |
| Jun 10, 2025 | 70.8 | GPQA Diamond | |
| May 28, 2025 | 81 | GPQA-Diamond (Pass@1) | |
| May 23, 2025 | Sarvam-M | 48 | GPQA Diamond (0.48) |
| May 22, 2025 | Claude Sonnet 4 | 83.8 | Graduate-level reasoning GPQA Diamond (parallel test-time compute) |
| May 22, 2025 | Claude Opus 4 | 83.3 | Graduate-level reasoning GPQA Diamond (parallel test-time compute) |
| May 20, 2025 | 49.7 | GPQA_Diamond | |
| May 11, 2025 | INTELLECT-2 | 66.8 | GPQA-Diamond |
| May 7, 2025 | 57.1 | GPQA Diamond 5-shot CoT | |
| Apr 30, 2025 | 65.8 | GPQA-D | |
| Apr 30, 2025 | 52 | GPQA Diamond | |
| Apr 30, 2025 | 57.1 | GPQA Diamond | |
| Apr 30, 2025 | 68.9 | GPQA-D | |
| Apr 29, 2025 | MiMo-7B-RL | 54.4 | GPQA Diamond (Pass@1) |
| Apr 29, 2025 | 71.1 | GPQA-Diamond | |
| Apr 17, 2025 | 78.3 | GPQA diamond single attempt (pass@1) | |
| Apr 16, 2025 | o4-mini | 81.4 | GPQA Diamond (no tools) |
| Apr 16, 2025 | o3 | 83.3 | GPQA Diamond PhD-Level Science Questions, o3 (no tools) |
| Apr 14, 2025 | GPT-4.1 nano | 50.3 | GPQA Diamond (model-extracted answers) |
| Apr 14, 2025 | GPT-4.1 mini | 65 | GPQA Diamond (model-extracted answers) |
| Apr 14, 2025 | GLM-Z1-32B-0414 | 66.1 | GPQA |
| Apr 10, 2025 | 77.3 | GPQA diamond | |
| Apr 8, 2025 | 60.6 | GPQA - Diamond (thinking) | |
| Apr 7, 2025 | 76 | GPQA (reasoning on) | |
| Apr 5, 2025 | 57.2 | GPQA Diamond | |
| Apr 5, 2025 | 69.8 | GPQA Diamond | |
| Apr 5, 2025 | 73.7 | GPQA Diamond | |
| Mar 27, 2025 | 30.8 | GPQA | |
| Mar 25, 2025 | 84 | GPQA diamond, single attempt (pass@1) | |
| Mar 24, 2025 | 59.1 | GPQA Diamond 59.1 (the original DeepSeek-V3 card. 68.4 is DeepSeek-V3-0324) | |
| Mar 21, 2025 | 69.3 | GPQA Diamond | |
| Mar 18, 2025 | 66.1 | GPQA Diamond (pass@1) | |
| Mar 18, 2025 | 66.7 | GPQA (reasoning on) | |
| Mar 17, 2025 | 46 | GPQA Diamond (5-shot CoT) | |
| Mar 16, 2025 | GPQA | ||
| Mar 12, 2025 | 42.4 | GPQA Diamond 0-shot (IT); PT GPQA 5-shot = 24.3 | |
| Mar 10, 2025 | Reka Flash 3 | 61.1 | GPQA |
| Feb 27, 2025 | 57.5 | GPQA-diamond | |
| Feb 27, 2025 | GPT-4.5 | 71.4 | GPQA (science) |
| Feb 26, 2025 | 30.4 | GPQA (0-shot, CoT), subset not stated | |
| Feb 24, 2025 | Claude 3.7 Sonnet | 84.8 | Graduate-level reasoning GPQA Diamond (64K extended thinking, parallel test-time compute) |
| Feb 19, 2025 | Grok-3 mini | 84 | GPQA (cons@64; pass@1 = 80.3) [Grok 3 mini Think] |
| Feb 19, 2025 | Grok 3 | 84.6 | GPQA (Graduate-Level Google-Proof Q&A (Diamond)) (cons@64; pass@1 = 80.2) |
| Feb 5, 2025 | 51.5 | GPQA (diamond) | |
| Feb 5, 2025 | 60.1 | GPQA (diamond) | |
| Feb 5, 2025 | 64.7 | GPQA (diamond) | |
| Jan 31, 2025 | o3-mini | 79.7 | PhD-level Science Questions (GPQA Diamond) |
| Jan 30, 2025 | 45.3 | gpqa_main_cot_5shot_instruct | |
| Jan 28, 2025 | GPQA-Diamond | ||
| Jan 21, 2025 | 74.2 | GPQA Diamond (Science) | |
| Jan 20, 2025 | 71.5 | GPQA-Diamond (Pass@1) | |
| Jan 15, 2025 | 54.4 | GPQA* (diamond) | |
| Jan 13, 2025 | 37.4 | GPQA-Diamond(0-shot) | |
| Dec 26, 2024 | 59.1 | GPQA-Diamond (Pass@1) | |
| Dec 17, 2024 | o1 | 75.7 | GPQA diamond |
| Dec 12, 2024 | 56.1 | GPQA (simple-evals) | |
| Dec 11, 2024 | 62.1 | GPQA (diamond) | |
| Dec 6, 2024 | 50.5 | GPQA Diamond (CoT), 0-shot | |
| Oct 22, 2024 | Claude 3.5 Sonnet (October 2024) | 65 | GPQA (Diamond) (0-shot CoT) |
| Sep 19, 2024 | 49 | GPQA | |
| Sep 12, 2024 | o1-preview | 78.3 | GPQA Diamond cons@64 |
| Aug 20, 2024 | 36.8 | GPQA (0-shot, CoT) | |
| Aug 13, 2024 | Grok-2 | 56 | GPQA |
| Jul 23, 2024 | 51.1 | GPQA (0-shot, CoT) | |
| Jul 18, 2024 | GPT-4o mini | 40.2 | GPQA |
| Jun 20, 2024 | Claude 3.5 Sonnet | 67.2 | GPQA (Diamond) (Maj@32 5-shot CoT) |
| Jun 7, 2024 | 42.4 | GPQA | |
| May 14, 2024 | 46.2 | GPQA, 0-shot | |
| May 13, 2024 | GPT-4o | 53.6 | GPQA (0-shot) |
| Apr 22, 2024 | GPQA (2-Shot; CoT) | ||
| Apr 18, 2024 | 39.5 | GPQA (0-shot) [instruct] | |
| Apr 11, 2024 | GPT-4 Turbo | 49.3 | GPQA% |
| Mar 4, 2024 | Claude 3 Opus | 59.5 | GPQA (Diamond) (Maj@32 5-shot CoT) |
| Feb 15, 2024 | 41.5 | GPQA, 0-shot (4-shot = 37.9%) |
168 system cards report it, from 32 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 20, 2026 | | 86.6 | GPQA Diamond | Capabilities |
| Nov 17, 2026 | GPT-6.2 Sol system card | 96 | GPQA Diamond (no tools), accuracy (Table 2, 2.1.1.1.1) | Bio and chem risk |
| Nov 17, 2026 | GPT-6.2 Astra system card | 97 | GPQA Diamond (no tools), accuracy (Table 2, 2.1.1.1.1) | Bio and chem risk |
| Oct 19, 2026 | | 83.5 | GPQA Diamond, high, sourced from Artificial Analysis, 85.6 xhigh (Tables 2 and 5) | Capabilities |
| Oct 15, 2026 | | 89.5 | GPQAD, mean of 8 repeats (Table 10) | Capabilities |
| Oct 15, 2026 | | 66.2 | GPQA Diamond, 5-shot CoT exact match, Table 6 | Capabilities |
| Oct 15, 2026 | | 76.5 | GPQA Diamond (no tools), BF16 acc. (Table 22) | Capabilities |
| Oct 15, 2026 | | 33.3 | GPQA Diamond, 5-shot CoT exact match, Table 6 | Capabilities |
| Oct 15, 2026 | | 86.4 | GPQA Diamond, released 24GB checkpoint (Table 7). BF16 88.83 | Capabilities |
| Oct 5, 2026 | Beam 1.0 Base technical report | 72.4 | GPQA Diamond, 0-shot temperature 0.7, Table 23 | Capabilities |
| Sep 22, 2026 | | 91 | GPQA Diamond (Table 2) | Capabilities |
| Sep 10, 2026 | | 90.9 | GPQA Diamond (Pass@1), Max effort (Table 3) | Capabilities |
| Aug 5, 2026 | | 82.2 | GPQA-Diamond (Table 6, Reasoning mode) | Capabilities |
| Jul 31, 2026 | | 73.2 | GPQA Diamond, TD, 64.6 no-think, 79.8 AR (Table 3) | Capabilities |
| Jul 30, 2026 | | 89.5 | GPQA Diamond (5. Evaluations table) | Capabilities |
| Jul 27, 2026 | Kimi K3 technical report | 93.5 | GPQA Diamond, max effort (Table 2) | Capabilities |
| Jul 24, 2026 | | 58.6 | GPQA Diamond, thinking (Table 5) | Capabilities |
| Jul 24, 2026 | | 43.4 | GPQA Diamond, thinking (Table 5) | Capabilities |
| Jul 24, 2026 | | 84.3 | GPQA Diamond, thinking (Table 5) | Capabilities |
| Jul 24, 2026 | | 82.3 | GPQA Diamond, thinking (Table 5) | Capabilities |
| Jul 22, 2026 | | 86.3 | GPQA-Diamond (Table 3) | Capabilities |
| Jul 15, 2026 | | 87.2 | GPQA Diamond (5. Evaluations table, effort 0.99) | Capabilities |
| Jun 13, 2026 | | 85.9 | GPQA-Diamond (Avg@16), xhigh (Table 6). High gives 76.10 | Capabilities |
| Jun 13, 2026 | | 76.2 | GPQA-Diamond, Mean@4, CoT (Table 5) | Capabilities |
| Jun 9, 2026 | | 87 | GPQA (no tools) (Table 10) | Capabilities |
| Jun 4, 2026 | | 50 | GPQA, 5-shot CoT EM | Capabilities |
| Jun 2, 2026 | | 84.2 | GPQA Diamond, pass@1 averaged over 16 rollouts (Table 11) | Capabilities |
| Jun 2, 2026 | | 84.6 | GPQA Diamond, Acc (Core reasoning table) | Capabilities |
| May 28, 2026 | Claude Opus 4.8 system card | 93.6 | GPQA Diamond, averaged over 25 trials (Table 8.1.A, 8.6) | Capabilities |
| May 26, 2026 | | 78 | GPQA-Diamond, MiniMax-M2 bar, printed label (Figure 9) | Capabilities |
| May 26, 2026 | | 89.8 | GPQA-Diamond, AA Index v4.0 protocol, pass@1 (Table 4) | Capabilities |
| May 26, 2026 | | 85.2 | GPQA-Diamond, AA Index v4.0 protocol, pass@1 (Table 4) | Capabilities |
| Apr 27, 2026 | | 72.2 | GPQA (no tools), GPQA-Diamond average of 4 runs (Table 10 and Section 4.4) | Capabilities |
| Apr 26, 2026 | | 90.1 | GPQA Diamond (Pass@1), DS-V4-Pro Max (Table 6) | Capabilities |
| Apr 26, 2026 | | 88.1 | GPQA Diamond (Pass@1), DeepSeek-V4-Flash Max (Table 7) | Capabilities |
| Apr 17, 2026 | | 83.9 | GPQA (Table 4) | Capabilities |
| Apr 16, 2026 | Claude Opus 4.7 system card | 94.2 | GPQA Diamond, average of 10 trials (Table 8.1.A, 8.4) | Capabilities |
| Apr 9, 2026 | | 80.5 | GPQA-Diamond (Table 3, Reasoning mode) | Capabilities |
| Apr 7, 2026 | Claude Mythos Preview system card | 94.5 | GPQA Diamond, adaptive thinking, max effort, average of 5 trials (Table 6.3.A; 94.55 in 6.6) | Capabilities |
| Mar 21, 2026 | | 83.8 | GPQA-Diamond (Table 1) | Capabilities |
| Mar 19, 2026 | | 76.1 | GPQA-Diamond (Table 1) | Capabilities |
| Mar 11, 2026 | | 79.2 | GPQA (no tools), 82.70 with tools (Table 5) | Capabilities |
| Mar 3, 2026 | | 86.9 | GPQA Diamond, no tools (results table, p.4) | Capabilities |
| Feb 28, 2026 | | 74.5 | GPQA, subset not stated (Table 8) | Capabilities |
| Feb 19, 2026 | | 94.3 | GPQA Diamond, no tools (results table, p.4) | Capabilities |
| Feb 19, 2026 | | 63.3 | GPQA Diamond, Trinity Large Preview (Table 4) | Capabilities |
| Feb 17, 2026 | GLM-5 technical report | 86 | GPQA-Diamond (Table 7) | Capabilities |
| Feb 17, 2026 | Claude Sonnet 4.6 system card | 89.9 | GPQA Diamond, adaptive thinking, max effort, average of 10 trials (Table 2.1.A, 2.9) | Capabilities |
| Feb 14, 2026 | | 88.9 | GPQA Diamond (Table 3) | Capabilities |
| Feb 11, 2026 | | 83.5 | GPQA-Diamond (Table 5). PaCoRe 85.0, with Python 84.4 (Table 12) | Capabilities |
| Feb 5, 2026 | Claude Opus 4.6 system card | 91.3 | GPQA Diamond, adaptive thinking, max effort, average of 5 trials (Table 2.3.A; 91.31 in 2.11) | Capabilities |
| Feb 4, 2026 | | 86.4 | GPQA-Diamond (Table 2) | Capabilities |
| Jan 30, 2026 | Kimi K2.5 technical report | 87.6 | GPQA-Diamond, Avg@8 (Table 4) | Capabilities |
| Jan 23, 2026 | | 85.2 | GPQA-Diamond (Avg@16), heavy mode (Table 2). Default mode 80.5 | Capabilities |
| Jan 14, 2026 | | 70.8 | GPQA-Diamond, SeRe (Tables 2 and 3). PaCoRe 73.99 | Capabilities |
| Jan 13, 2026 | | 71.2 | GPQA Diamond, reported as pass@16 (Table 5) | Capabilities |
| Jan 11, 2026 | | 68.1 | GPQA-Diamond (Table 6) | Capabilities |
| Jan 6, 2026 | MiMo-V2-Flash technical report | 84.3 | GPQA-Diamond (Table 9). The launch model card prints 83.7 | Capabilities |
| Jan 5, 2026 | | 61.3 | GPQA-Diamond (Table 6). DeepConf@512 70.2 (Table 7) | Capabilities |
| Jan 5, 2026 | | 79.1 | GPQA-Diamond (Table 3, Reasoning mode). Non-reasoning 70.6 (Table 4) | Capabilities |
| Dec 23, 2025 | | 73 | GPQA (no tools) (Table 3). With tools 75.00 | Capabilities |
| Dec 18, 2025 | INTELLECT-3 technical report | 74.4 | GPQA (Diamond subset), Avg@4 (Table 2, Appendix A.1) | Capabilities |
| Dec 15, 2025 | | 58.1 | GPQA, Final Think 3.0 column (Table 14) | Capabilities |
| Dec 15, 2025 | | 57.5 | GPQA (Table 1). Table 14 prints 56.7 for Final Think 3.1 | Capabilities |
| Dec 2, 2025 | | 79.6 | GPQA Diamond acc (Table 1) | Capabilities |
| Dec 2, 2025 | | 81.4 | GPQA Diamond acc (Table 1) | Capabilities |
| Dec 1, 2025 | | 82.4 | GPQA Diamond (Pass@1), DeepSeek-V3.2 Thinking (Table 2) | Capabilities |
| Dec 1, 2025 | | 85.7 | GPQA Diamond (Pass@1), DeepSeek-V3.2-Speciale (Table 3) | Capabilities |
| Nov 28, 2025 | LFM2-8B-A1B technical report | 29.3 | GPQA, LFM2-8B-A1B (Table 7) | Capabilities |
| Nov 28, 2025 | LFM2-1.2B technical report | 31.5 | GPQA, LFM2-1.2B (Table 6) | Capabilities |
| Nov 26, 2025 | | 77.1 | GPQA (Table 6) | Capabilities |
| Nov 24, 2025 | Claude Opus 4.5 system card | 87 | GPQA Diamond, 64k thinking, avg of 5 trials (Table 2.3.A; 86.95% in 2.18) | Capabilities |
| Nov 6, 2025 | | 64.1 | GPQA, reasoning on, SFT stage 4 (Table 6), GPQA-Diamond in 4.2 | Capabilities |
| Oct 31, 2025 | | 74.4 | GPQA-diamond (acc) (Table 13) | Capabilities |
| Oct 21, 2025 | | 78.6 | GPQA-Diamond (Table 1) | Capabilities |
| Sep 30, 2025 | | 40.6 | GPQA, without reasoning, subset not stated (Table 13) | Capabilities |
| Sep 29, 2025 | | 79.9 | GPQA-Diamond (Pass@1), DeepSeek-V3.2-Exp (Table 1) | Capabilities |
| Sep 26, 2025 | | 80.8 | GPQA diamond, single attempt (pass@1) (Table, p.4) | Capabilities |
| Sep 25, 2025 | | 83.3 | GPQA (Table 20, thinking on). Thinking off 81.3 | Capabilities |
| Sep 23, 2025 | | 81.5 | GPQA-Diamond(Mean@16), LLM judge (Table 2) | Capabilities |
| Sep 22, 2025 | | 69.6 | GPQA, Instruct (Table 4). Thinking 73.1 (Table 5) | Capabilities |
| Sep 1, 2025 | | 73.2 | GPQA-diamond (acc) (Table 3) | Capabilities |
| Aug 25, 2025 | | 73.2 | GPQA, cited without the Diamond label (Table 2) | Capabilities |
| Aug 25, 2025 | Hermes 4 405B technical report | 70.6 | GPQA Diamond, reasoning mode (Table 3). Non-reasoning 39.4. v1 printed 70.5 | Capabilities |
| Aug 21, 2025 | | 77.3 | GPQA, Diamond subset (Table 2) | Capabilities |
| Aug 20, 2025 | | 64 | GPQA-D, NVIDIA-Nemotron-Nano-9B-v2 (Figure 1) | Capabilities |
| Aug 8, 2025 | GLM-4.5 technical report | 79.1 | GPQA, Avg@8 (Table 4) | Capabilities |
| Aug 8, 2025 | GLM-4.5-Air technical report | 75 | GPQA, Avg@8 (Table 4) | Capabilities |
| Aug 5, 2025 | gpt-oss-20b model card | 71.5 | GPQA Diamond (no tools), high reasoning (with tools 74.2) (Table 3) | Capabilities |
| Aug 5, 2025 | gpt-oss-120b model card | 80.1 | GPQA Diamond (no tools), high reasoning (with tools 80.9) (Table 3) | Capabilities |
| Jul 30, 2025 | | 49.7 | GPQA_Diamond, average accuracy over 3 repetitions (Table 24) | Capabilities |
| Jul 21, 2025 | Kimi K2 technical report | 75.1 | GPQA-Diamond (Avg@8) (Table 3) | Capabilities |
| Jul 15, 2025 | | 75.4 | GPQA-Diamond, n=8 (Table 3, Reasoning). Non-reasoning 63.7 (Table 4) | Capabilities |
| Jul 9, 2025 | | 45.1 | GPQA Diamond, pass@1 averaged over 8 samples (Table 4) | Capabilities |
| Jul 7, 2025 | | 86.4 | GPQA (diamond), single attempt pass@1 (Table 3) | Capabilities |
| Jul 7, 2025 | | 82.8 | GPQA (diamond), single attempt pass@1 (Table 3) | Capabilities |
| Jun 26, 2025 | | 71.2 | GPQA-Diamond, slow-thinking (Table 3). Fast-thinking 61.8 (Table 4) | Capabilities |
| Jun 16, 2025 | | 70 | GPQA Diamond, avg of 32 samples (Table 2) | Capabilities |
| Jun 12, 2025 | | 68.2 | GPQA, SFT + RL (Magistral Small) column (Table 3) | Capabilities |
| Jun 12, 2025 | | 70.8 | GPQA (Table 2) | Capabilities |
| May 14, 2025 | | 62 | GPQA-Diamond, thinking mode (Table 17). Non-thinking 39.3 (Table 18) | Capabilities |
| May 14, 2025 | | 55.9 | GPQA-Diamond, thinking mode (Table 17). Non-thinking 41.7 (Table 18) | Capabilities |
| May 14, 2025 | | 68.4 | GPQA-Diamond, thinking mode (Table 13). Non-thinking 54.6 (Table 14) | Capabilities |
| May 14, 2025 | | 65.8 | GPQA-Diamond, thinking mode (Table 15). Non-thinking 54.8 (Table 16) | Capabilities |
| May 14, 2025 | | 71.1 | GPQA-Diamond, thinking mode (Table 11). Non-thinking 62.9 (Table 12) | Capabilities |
| May 14, 2025 | | 64 | GPQA-Diamond, thinking mode (Table 15). Non-thinking 54.8 (Table 16) | Capabilities |
| May 14, 2025 | | 40.1 | GPQA-Diamond, thinking mode (Table 19). Non-thinking 28.6 (Table 20) | Capabilities |
| May 12, 2025 | MiMo-7B-RL technical report | 54.4 | GPQA Diamond (Pass@1), average of 8 runs (Table 4, Table 6) | Capabilities |
| May 12, 2025 | INTELLECT-2 technical report | 66.8 | GPQA-Diamond, evalchemy defaults, target length 10,000 tokens (Table 1) | Capabilities |
| May 2, 2025 | | 66.7 | GPQA-Diamond, reasoning on, reasoning off 50.0 (Table 4) | Capabilities |
| May 2, 2025 | | 76 | GPQA-Diamond, reasoning on, reasoning off 56.6 (Table 5) | Capabilities |
| Apr 30, 2025 | | 67.1 | GPQA-D, average pass@1 (Table 1) | Capabilities |
| Apr 30, 2025 | | 69.3 | GPQA-D, average pass@1 (Table 1) | Capabilities |
| Apr 30, 2025 | | 52 | GPQA Diamond, pass@1 (Table 3) | Capabilities |
| Apr 30, 2025 | | 57.1 | GPQA Diamond, 0-shot CoT, +/-6.9 (Table 1) | Capabilities |
| Apr 29, 2025 | | 78.3 | GPQA diamond, single attempt (pass@1) (Table, p.4) | Capabilities |
| Apr 28, 2025 | | 84 | GPQA diamond, single attempt pass@1 (Table, p.4, column Gemini 2.5 Pro Experimental 03-25) | Capabilities |
| Apr 15, 2025 | | 60.1 | GPQA (diamond) (Table, p.3, column Gemini 2.0 Flash (GA)) | Capabilities |
| Apr 10, 2025 | | 51.5 | GPQA (diamond) (Table, p.3) | Capabilities |
| Apr 10, 2025 | | 77.3 | GPQA diamond, mean of 8 samples (Table 2) | Capabilities |
| Apr 5, 2025 | | 57.2 | GPQA Diamond, 0-shot, accuracy, instruction tuned (Instruction tuned models table) | Capabilities |
| Apr 5, 2025 | | 69.8 | GPQA Diamond, 0-shot, accuracy, instruction tuned (Instruction tuned models table) | Capabilities |
| Apr 1, 2025 | | 50.8 | GPQA, Diamond subset, 0-shot CoT (Table 3, Table 15) | Capabilities |
| Mar 26, 2025 | | 30.8 | GPQA (Table 1), read as GPQA Diamond as in the launch record | Capabilities |
| Mar 25, 2025 | | 30.8 | GPQA Diamond, IT, zero-shot (Table 6) | Capabilities |
| Mar 25, 2025 | | 42.4 | GPQA Diamond, IT, zero-shot (Table 6) | Capabilities |
| Mar 25, 2025 | | 19.2 | GPQA Diamond, IT, zero-shot (Table 6) | Capabilities |
| Mar 25, 2025 | | 40.9 | GPQA Diamond, IT, zero-shot (Table 6) | Capabilities |
| Mar 16, 2025 | | 66.1 | GPQA Diamond, pass@1 with k=16 (Table 3) | Capabilities |
| Mar 3, 2025 | | 30.4 | GPQA (0-Shot, CoT), subset not stated (Table 7). GPQA Diamond 36.9 in the CoT reasoning comparison (Table 9) | Capabilities |
| Feb 19, 2025 | | 49 | GPQA (Table 4) | Capabilities |
| Jan 26, 2025 | | 39.9 | GPQA (Table 6), mapped to GPQA Diamond as the almanac alias does | Capabilities |
| Jan 20, 2025 | | 71.5 | GPQA Diamond (Pass@1), DeepSeek-R1 (Table 8, same as final column of Table 3) | Capabilities |
| Jan 20, 2025 | | 62.1 | GPQA Diamond pass@1, DeepSeek-R1-Distill-Qwen-32B (Table 15) | Capabilities |
| Jan 20, 2025 | | 59.1 | GPQA Diamond pass@1, DeepSeek-R1-Distill-Qwen-14B (Table 15) | Capabilities |
| Jan 20, 2025 | | 33.8 | GPQA Diamond pass@1, DeepSeek-R1-Distill-Qwen-1.5B (Table 15) | Capabilities |
| Jan 20, 2025 | | 65.2 | GPQA Diamond pass@1, DeepSeek-R1-Distill-Llama-70B (Table 15) | Capabilities |
| Jan 14, 2025 | | 54.4 | GPQA (diamond), 0-shot CoT (Table 8) | Capabilities |
| Dec 31, 2024 | | 35.9 | GPQA, unseen suite, 0-shot CoT, subset not stated (Table 24) | Capabilities |
| Dec 26, 2024 | | 59.1 | GPQA-Diamond (Pass@1), DeepSeek-V3 (Table 6) | Capabilities |
| Dec 19, 2024 | | 36.4 | GPQA, variant not stated, Qwen2.5-7B-Instruct (Table 8) | Capabilities |
| Dec 19, 2024 | | 49 | GPQA, variant not stated, Qwen2.5-72B-Instruct (Table 6) | Capabilities |
| Dec 19, 2024 | | 49.5 | GPQA, variant not stated, Qwen2.5-32B-Instruct (Table 7) | Capabilities |
| Dec 19, 2024 | | 45.5 | GPQA, variant not stated, Qwen2.5-14B-Instruct (Table 7) | Capabilities |
| Dec 19, 2024 | | 42.3 | GPQA, variant not stated, Qwen2.5-Turbo (Table 7) | Capabilities |
| Dec 12, 2024 | | 56.1 | GPQA, simple-evals diamond set (Table 1) | Capabilities |
| Dec 6, 2024 | | 50.5 | GPQA Diamond (CoT), 0-shot, acc (Instruction tuned models table) | Capabilities |
| Nov 23, 2024 | | 32.8 | GPQA (0-shot, CoT), subset not stated, post-trained (Table 2) | Capabilities |
| Nov 23, 2024 | | 46.7 | GPQA (0-shot, CoT), subset not stated, post-trained (Table 2) | Capabilities |
| Nov 23, 2024 | | 51.1 | GPQA (0-shot, CoT), subset not stated, post-trained (Table 2) | Capabilities |
| Oct 22, 2024 | Claude 3.5 Sonnet (October 2024) model card | 65 | GPQA (Diamond), 0-shot CoT (Table 7) | Capabilities |
| Oct 22, 2024 | Claude 3.5 Haiku model card | 41.6 | GPQA (Diamond), 0-shot CoT (Table 8) | Capabilities |
| Aug 30, 2024 | | 34.3 | GPQA (2-Shot, CoT), subset not stated (Section 3 table) | Capabilities |
| Aug 30, 2024 | | 32.8 | GPQA (2-Shot, CoT), subset not stated (Section 3 table) | Capabilities |
| Aug 30, 2024 | | 36.8 | GPQA (0-shot, CoT), subset not stated (Table 3) | Capabilities |
| Aug 30, 2024 | | 27.2 | GPQA (0-shot, CoT), subset not stated (Table 3) | Capabilities |
| Jul 23, 2024 | | 30.4 | GPQA, 0-shot, em, subset not stated, instruction tuned (Instruction tuned models table) | Capabilities |
| Jul 23, 2024 | | 46.7 | GPQA, 0-shot, em, subset not stated, instruction tuned (Instruction tuned models table) | Capabilities |
| Jul 23, 2024 | | 50.7 | GPQA, 0-shot, em, subset not stated, instruction tuned (Instruction tuned models table) | Capabilities |
| Jul 15, 2024 | | 42.4 | GPQA, variant not stated, Qwen2-72B-Instruct (Table 6) | Capabilities |
| Jun 20, 2024 | Claude 3.5 Sonnet model card | 59.4 | GPQA (Diamond), 0-shot CoT. Maj@32 5-shot CoT 67.2 (Table 1) | Capabilities |
| Jun 18, 2024 | GLM-4 technical report | 35.7 | GLM-4 (0116), GPQA, subset not stated (Table 2) | Capabilities |
| Jun 18, 2024 | ChatGLM3-6B technical report | 26.8 | ChatGLM3-6B-Base, GPQA, subset not stated (Table 1) | Capabilities |
| Apr 18, 2024 | | 34.2 | GPQA (0-shot), subset not stated, instruction tuned (Instruction tuned models table) | Capabilities |
| Apr 18, 2024 | | 39.5 | GPQA (0-shot), subset not stated, instruction tuned (Instruction tuned models table) | Capabilities |
| Mar 4, 2024 | Claude 3 Sonnet model card | 40.4 | GPQA (Diamond), 0-shot CoT, mean of 10 rollouts. 5-shot CoT 42.9, Maj@32 5-shot CoT 46.3 (Table 1, Table 8) | Capabilities |
| Mar 4, 2024 | Claude 3 Opus model card | 50.4 | GPQA (Diamond), 0-shot CoT, mean of 10 rollouts. 5-shot CoT 53.3, Maj@32 5-shot CoT 59.5 (Table 1, Table 8) | Capabilities |
| Mar 4, 2024 | Claude 3 Haiku model card | 33.3 | GPQA (Diamond), 0-shot CoT, mean of 10 rollouts. 5-shot CoT 36.4, Maj@32 5-shot CoT 40.1 (Table 1, Table 8) | Capabilities |
The lab's number comes out ahead in 94 of 123 pairs that differ
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 6 of them match, which usually means one side copied the other. Every benchmark.
| Model | Lab says | Board says | Gap | What differs |
|---|---|---|---|---|
| o1-preview | 78.3% GPQA Diamond cons@64 | 50.3% Epoch AI run | +28 pts | The lab reports consensus across 64 responses, while Epoch averages scores from individual responses. |
| | 70.1% GPQA Diamond | 47.6% Epoch AI run | +22 pts | |
| | 74.2% GPQA Diamond (Science) | 57.1% Epoch AI run | +17 pts | |
| glm-4.7-flash | 75.2% GPQA | 60.5% Epoch AI run | +15 pts | |
| | 59.1% GPQA Diamond pass@1, DeepSeek-R1-Distill-Qwen-14B (Table 15) | 44.7% Epoch AI run | +14 pts | |
| GPT-5.6 Terra | 92.9% GPQA Diamond | 79.3% Leaderboard | +14 pts | |
| | 68.2% GPQA, SFT + RL (Magistral Small) column (Table 3) | 56.1% Epoch AI run | +12 pts | |
| GPT-5.6 Luna | 92.3% GPQA Diamond | 80.8% Leaderboard | +11 pts | |
| gpt-oss-20b | 71.5% GPQA Diamond (without tools) | 60.8% Epoch AI run, medium | +11 pts | OpenAI reports its evaluation at high reasoning effort without tools. The paired Epoch score uses medium reasoning effort. |
| Claude 3.5 Sonnet (October 2024) | 65% GPQA (Diamond) (0-shot CoT) | 55.3% Epoch AI run | +9.7 pts | |
| | 65.2% GPQA Diamond pass@1, DeepSeek-R1-Distill-Llama-70B (Table 15) | 55.7% Epoch AI run | +9.5 pts | |
| Grok 3 | 84.6% GPQA (Graduate-Level Google-Proof Q&A (Diamond)) (cons@64; pass@1 = 80.2) | 75.8% Epoch AI run | +8.8 pts | |
| | 59.1% GPQA Diamond 59.1 (the original DeepSeek-V3 card. 68.4 is DeepSeek-V3-0324) | 67.6% Epoch AI run | −8.5 pts | |
| | 34.2% GPQA (0-shot), subset not stated, instruction tuned (Instruction tuned models table) | 26.1% Epoch AI run | +8.1 pts | |
| Grok-3 mini | 84% GPQA (cons@64; pass@1 = 80.3) [Grok 3 mini Think] | 76.3% Epoch AI run, low | +7.7 pts | |
| | 30.8% GPQA Diamond, IT, zero-shot (Table 6) | 23.2% Epoch AI run | +7.6 pts | |
| GPT-5 mini | 82.3% GPQA diamond (no tools) | 75% Epoch AI run, high | +7.3 pts | |
| Claude Opus 4 | 83.3% Graduate-level reasoning GPQA Diamond (parallel test-time compute) | 76.3% Epoch AI run, 16K | +7.0 pts | |
| granite-4.0-1b | 29.9% GPQA (0-shot, CoT; subset not stated) | 24% Epoch AI run | +5.9 pts | |
| | 32.8% GPQA (0-shot, CoT), subset not stated, post-trained (Table 2) | 27% Epoch AI run | +5.8 pts | |
| | 57.2% GPQA Diamond | 51.8% Epoch AI run | +5.4 pts | |
| Claude 3.5 Sonnet | 59.4% GPQA (Diamond), 0-shot CoT. Maj@32 5-shot CoT 67.2 (Table 1) | 54% Epoch AI run, Jun 2024 | +5.4 pts | |
| | 42.4% GPQA Diamond 0-shot (IT); PT GPQA 5-shot = 24.3 | 47.7% Epoch AI run | −5.3 pts | |
| | 62% GPQA-Diamond, thinking mode (Table 17). Non-thinking 39.3 (Table 18) | 56.8% Epoch AI run | +5.2 pts | |
| | 86.9% GPQA Diamond (No tools) | 81.8% Epoch AI run, high | +5.1 pts | |
| Claude 3.7 Sonnet | 84.8% Graduate-level reasoning GPQA Diamond (64K extended thinking, parallel test-time compute) | 79.7% Epoch AI run, 64k thinking | +5.1 pts | |
| GPT-4o May 2024 | 53.6% GPQA (0-shot) | 48.9% Epoch AI run | +4.7 pts | |
| | 81% GPQA-Diamond (Pass@1) | 76.3% Epoch AI run | +4.7 pts | |
| Claude Sonnet 4 | 83.8% Graduate-level reasoning GPQA Diamond (parallel test-time compute) | 79.2% Epoch AI run, 59K | +4.6 pts | |
| gpt-oss-120b | 80.1% GPQA Diamond (without tools) | 75.8% Epoch AI run, high | +4.3 pts | |
| GPT-5.4 Nano | 82.8% GPQA Diamond | 78.5% Epoch AI run, high | +4.3 pts | |
| | 65.8% GPQA-Diamond, thinking mode (Table 15). Non-thinking 54.8 (Table 16) | 61.7% Epoch AI run | +4.1 pts | |
| Claude Opus 4.7 | 94.2% Graduate-level reasoning GPQA Diamond | 90.2% Epoch AI run, xhigh | +4.0 pts | |
| | 60.1% GPQA (diamond) | 64.1% Epoch AI run | −4.0 pts | |
| GLM-5.1 | 86.2% GPQA-Diamond | 89.9% Epoch AI run | −3.7 pts | |
| Claude Opus 4.1 | 80.9% Graduate-level reasoning GPQA Diamond | 77.3% Epoch AI run, 16K | +3.6 pts | |
| | 55.9% GPQA-Diamond, thinking mode (Table 17). Non-thinking 41.7 (Table 18) | 52.3% Epoch AI run | +3.6 pts | |
| Claude 3.5 Haiku | 41.6% GPQA (Diamond), 0-shot CoT (Table 8) | 38.1% Epoch AI run, Oct 2024 | +3.5 pts | |
| | 49.5% GPQA, variant not stated, Qwen2.5-32B-Instruct (Table 7) | 46.1% Epoch AI run | +3.4 pts | |
| Claude 3 Opus | 50.4% GPQA (Diamond), 0-shot CoT, mean of 10 rollouts. 5-shot CoT 53.3, Maj@32 5-shot CoT 59.5 (Table 1, Table 8) | 47.2% Epoch AI run | +3.2 pts | |
| GPT-5 | 89.4% GPQA Diamond, GPT-5 pro (python) | 86.2% Epoch AI run, high | +3.2 pts | |
| GPT-5.5 Instant | 85.6% GPQA (PhD-Level Science) | 82.5% Epoch AI run | +3.1 pts | |
| Claude 3 Haiku | 33.3% GPQA (Diamond), 0-shot CoT, mean of 10 rollouts. 5-shot CoT 36.4, Maj@32 5-shot CoT 40.1 (Table 1, Table 8) | 36.3% Epoch AI run | −3.0 pts | |
| | 46.1% GPQA Diamond (5-shot CoT) | 49% Epoch AI run | −2.9 pts | |
| | 69.8% GPQA Diamond | 67% Epoch AI run, FP8 | +2.8 pts | |
| | 81.7% GPQA Diamond | 79% Epoch AI run | +2.7 pts | |
| GPT-4.5 | 71.4% GPQA (science) | 68.7% Epoch AI run | +2.7 pts | |
| GPT-4 Turbo Apr 2024 | 49.3% GPQA% | 46.6% Epoch AI run | +2.7 pts | |
| o3-mini | 79.7% PhD-level Science Questions (GPQA Diamond) | 77% Epoch AI run, high | +2.7 pts | |
| | 68.4% GPQA-Diamond, thinking mode (Table 13). Non-thinking 54.6 (Table 14) | 65.7% Epoch AI run | +2.7 pts | |
| | 59.1% GPQA-Diamond (Pass@1) | 56.5% Epoch AI run | +2.6 pts | |
| Claude Opus 4.8 | 93.6% GPQA Diamond | 91% Epoch AI run, max | +2.6 pts | |
| Claude Sonnet 4.6 | 89.9% Graduate-level reasoning GPQA Diamond | 87.4% Epoch AI run, 32k thinking | +2.5 pts | |
| | 46.7% GPQA, 0-shot, em, subset not stated, instruction tuned (Instruction tuned models table) | 44.2% Epoch AI run | +2.5 pts | |
| GPT-4o mini | 40.2% GPQA | 37.7% Epoch AI run | +2.5 pts | |
| | 57.1% GPQA Diamond 5-shot CoT | 59.5% Epoch AI run | −2.4 pts | |
| | 90.3% GPQA Diamond | 87.9% Epoch AI run | +2.4 pts | |
| GLM-4.7 | 85.7% GPQA-Diamond | 83.3% Epoch AI run | +2.4 pts | |
| | 40.1% GPQA-Diamond, thinking mode (Table 19). Non-thinking 28.6 (Table 20) | 38% Epoch AI run | +2.1 pts | |
| | 88.4% GPQA | 86.4% Epoch AI run, none | +2.0 pts | |
| | 62.1% GPQA Diamond pass@1, DeepSeek-R1-Distill-Qwen-32B (Table 15) | 64.1% Epoch AI run | −2.0 pts | |
| | 90.4% GPQA | 88.4% Epoch AI run, 2026-04-02 | +2.0 pts | |
| | 45.3% gpqa_main_cot_5shot_instruct | 47.3% Epoch AI run | −2.0 pts | |
| | 87.8% GPQA Diamond | 85.9% Epoch AI run | +1.9 pts | |
| GLM-5 | 86% GPQA-Diamond | 87.8% Epoch AI run | −1.8 pts | |
| o4-mini | 81.4% GPQA Diamond (no tools) | 79.6% Epoch AI run, high | +1.8 pts | |
| Claude Haiku 4.5 | 73% Graduate-level reasoning GPQA Diamond | 71.2% Epoch AI run, 32K | +1.8 pts | |
| GPT-5 nano | 71.2% GPQA diamond (no tools) | 69.4% Epoch AI run, high | +1.8 pts | |
| | 87% GPQA (no tools) | 85.3% Epoch AI run | +1.7 pts | |
| | 42.4% GPQA | 40.8% Epoch AI run | +1.6 pts | |
| | 46% GPQA Diamond (5-shot CoT) | 47.5% Epoch AI run | −1.5 pts | |
| | 92.4% GPQA Diamond | 90.9% Epoch AI run | +1.5 pts | |
| o3 | 83.3% GPQA Diamond PhD-Level Science Questions, o3 (no tools) | 81.8% Epoch AI run, high | +1.5 pts | |
| | 84.3% GPQA Diamond | 82.8% Leaderboard | +1.5 pts | |
| | 40.9% GPQA Diamond, IT, zero-shot (Table 6) | 39.5% Epoch AI run | +1.4 pts | |
| GPT-4.1 nano | 50.3% GPQA Diamond (model-extracted answers) | 48.9% Epoch AI run | +1.4 pts | |
| | 86% GPQA | 84.8% Epoch AI run, none | +1.2 pts | |
| GPT-5.4 Mini | 88% GPQA Diamond | 86.9% Epoch AI run, xhigh | +1.1 pts | |
| | 86.4% GPQA diamond, single attempt (pass@1) | 85.3% Epoch AI run | +1.1 pts | |
| GPT-5.6 Sol | 94.6% GPQA Diamond | 93.5% Epoch AI run, max | +1.1 pts | |
| Claude Sonnet 4.5 | 83.4% Graduate-level reasoning GPQA Diamond | 82.3% Epoch AI run, 59k thinking | +1.1 pts | |
| o1 | 75.7% GPQA diamond | 76.8% Epoch AI run, high | −1.1 pts | |
| | 39.5% GPQA (0-shot) [instruct] | 40.6% Epoch AI run | −1.1 pts | |
| | 87.2% GPQA Diamond (Artificial Analysis) | 88.3% Epoch AI run, xhigh | −1.1 pts | |
| | 90.9% GPQA Diamond (Pass@1) | 89.8% Epoch AI run, max | +1.1 pts | |
| | 81.1% GPQA | 80% Epoch AI run | +1.0 pts | |
| GPT-6 Astra | 96% GPQA Diamond | 95% Leaderboard | +1.0 pts | |
| | 82.4% GPQA Diamond (Pass@1) | 83.4% Epoch AI run, Thinking | −1.0 pts | |
| | 90.4% GPQA Diamond (No tools) | 89.4% Epoch AI run, high | +1.0 pts | |
| GPT-5.2 | 92.4% GPQA Diamond (no tools) Science questions | 91.4% Epoch AI run, xhigh | +1.0 pts | |
| | 89.5% GPQA Diamond | 88.5% Epoch AI run, xhigh | +1.0 pts | |
| | 64.7% GPQA (diamond) | 65.7% Epoch AI run | −1.0 pts | |
| Claude Opus 4.5 | 87% Graduate-level reasoning GPQA Diamond | 86% Epoch AI run, 32k thinking | +1.0 pts | |
| | 82.3% GPQA Diamond | 83.2% Leaderboard | −0.9 pts | |
| | 36.4% GPQA, variant not stated, Qwen2.5-7B-Instruct (Table 8) | 35.5% Epoch AI run | +0.9 pts | |
| GPT-4.1 mini | 65% GPQA Diamond (model-extracted answers) | 65.8% Epoch AI run | −0.8 pts | |
| | 90.1% GPQA Diamond (Pass@1) | 90.9% Epoch AI run, high | −0.8 pts | |
| Claude Opus 4.6 | 91.3% Graduate-level reasoning GPQA Diamond | 90.5% Epoch AI run, 32k thinking | +0.8 pts | |
| | 19.2% GPQA Diamond, IT, zero-shot (Table 6) | 19.9% Epoch AI run | −0.8 pts | |
| | 84.2% GPQA Diamond | 83.5% Epoch AI run | +0.7 pts | |
| | 91.9% GPQA Diamond (No tools) | 92.6% Epoch AI run | −0.7 pts | |
| GLM-5.2 | 91.2% GPQA-Diamond | 91.9% Epoch AI run, max | −0.7 pts | |
| | 42.3% GPQA, variant not stated, Qwen2.5-Turbo (Table 7) | 41.8% Epoch AI run | +0.5 pts | |
| Grok 4 | 87.5% GPQA (Science), no tools | 87% Epoch AI run | +0.5 pts | |
| GPT-5.4 | 92.8% GPQA Diamond (GPT-5.4) | 93.3% Epoch AI run, xhigh | −0.5 pts | |
| GPT-5.1 | 88.1% GPQA Diamond (no tools) | 87.6% Epoch AI run, high | +0.5 pts | |
| | 89% GPQA Diamond | 88.6% Epoch AI run, xhigh | +0.4 pts | |
| GPT-5.5 | 93.6% GPQA Diamond | 94% Epoch AI run, xhigh | −0.4 pts | |
| | 71.1% GPQA-Diamond | 70.7% Epoch AI run | +0.4 pts | |
| Kimi K3 | 93.5% GPQA Diamond | 93.1% Epoch AI run, max | +0.4 pts | |
| | 94.3% GPQA Diamond (No tools) | 93.9% Leaderboard | +0.4 pts | |
| | 46.2% GPQA, 0-shot | 45.9% Epoch AI run | +0.3 pts | |
| | 89.5% GPQA Diamond (PhD Level Reasoning) | 89.8% Epoch AI run | −0.3 pts | |
| Kimi K2 Thinking | 84.5% GPQA, no tools | 84.2% Epoch AI run | +0.3 pts | |
| Kimi K2.6 | 90.5% GPQA-Diamond | 90.8% Epoch AI run | −0.3 pts | |
| | 64% GPQA-Diamond, thinking mode (Table 15). Non-thinking 54.8 (Table 16) | 63.8% Epoch AI run | +0.2 pts | |
| | 71.5% GPQA-Diamond (Pass@1) | 71.7% Epoch AI run | −0.2 pts | |
| | 33.8% GPQA Diamond pass@1, DeepSeek-R1-Distill-Qwen-1.5B (Table 15) | 33.6% Epoch AI run | +0.2 pts | |
| GPT-5.4 Pro | 94.4% GPQA Diamond | 94.6% Epoch AI run, xhigh | −0.2 pts | |
| Claude 3 Sonnet | 40.4% GPQA (Diamond), 0-shot CoT, mean of 10 rollouts. 5-shot CoT 42.9, Maj@32 5-shot CoT 46.3 (Table 1, Table 8) | 40.6% Epoch AI run | −0.2 pts | |
| | 51.1% GPQA (0-shot, CoT) | 50.9% Epoch AI run | +0.2 pts | |
| | 84% GPQA diamond, single attempt (pass@1) | 83.8% Epoch AI run | +0.2 pts | |
| GPT-6.2 Astra | 97% GPQA Diamond (no tools), accuracy (Table 2, 2.1.1.1.1) | 96.8% Epoch AI run, max | +0.2 pts | |
| | 49% GPQA | 49.1% Epoch AI run | Same | |
| | 71.4% GPQA-D | 71.5% Epoch AI run | Same | |
| | 92.6% GPQA Diamond | 92.7% Epoch AI run, xhigh | Same | |
| | 56.1% GPQA (simple-evals) | 56.1% Epoch AI run | Same | |
| Kimi K3.1 Pro | 94.2% GPQA-Diamond | 94.2% Epoch AI run, Together, max | Same | |
| Kimi K2.5 | 87.6% GPQA-Diamond | 87.6% Epoch AI run, Fireworks | Same |
More in Science and reasoning
- Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87%
- HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels.
Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%
- CritPt v2 In use Models solve unpublished research-level physics challenges written by working physicists, with answers graded automatically. The v2 release revised the public test challenges. GPT-6 Astra 83.1%
Data audit
10 of 11 checks pass, 1 warning. Look first at the Lab claims check. 7 of 129 pairs differ by more than 10 points, the largest for Magistral Small 1.2 (+22 pts). Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (20 entries), Epoch AI’s runs (231 models), launch posts (254 scores) and system cards (168 scores).
- Pass Models scored 408 models we track.
- Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
- Pass Frontier models 9 of the top 10 have a score, led by GPT-6.2 Astra (#1).
- Note Headroom Saturated. The best is 97%, past 85% since June 2025.
Leaderboard import 3 pass
- Pass Importer We read the board on kaggle.com: 20 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 warning 1 pass
- Warning Lab claims 7 of 129 pairs differ by more than 10 points, the largest for Magistral Small 1.2 (+22 pts).
- Pass Known issues No published audit found problems.