GPQA Diamond (Epoch AI run) leaderboard
- Gemini 4 Argon leads
- 96.4%
- models measured
- 117
- lead over #2
- +0.6 pts
- median
- 85.9%
Leader on GPQA Diamond (Epoch AI run) · read October 22
On October 22, 2026, 133 of our catalog's models had a GPQA Diamond (Epoch AI run) result, and Google Gemini 4 Argon led with 96.4%, ahead of OpenAI GPT-6 Astra at 95.8%.
The same GPQA Diamond questions as the Vals AI board, run by Epoch AI itself: graduate-level science questions written to resist web search. Share answered correctly.
Role in the index
GPQA Diamond (Epoch AI run) carries 30% of the Reasoning Index. A model missing it is scored on the other benchmarks of that index. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | Gemini 4 ArgonGoogle | 96.4% | 81.0 | $10.00 | |
| 2 | GPT-6 AstraOpenAI | 95.8% | 85.1 | $50.00 | |
| 3 | Claude Sonnet 5.5Anthropic | 95.6% | 87.3 | $10.00 | |
| 4 | Gemini 3.8 FlashGoogle | 95.4% | 61.6 | $3.75 | |
| 5 | GPT-6.1 SolOpenAI | 95.4% | 82.4 | $10.00 | |
| 6 | Gemini 3.7 FlashGoogle | 94.8% | 58.0 | $3.75 | |
| 7 | GPT-5.4 ProOpenAI | 94.6% | Not benchmarked | $180.00 | |
| 8 | Gemini 3.1 Pro PreviewGoogle | 94.4% | 32.3 | $12.00 | |
| 9 | GPT-6 SolOpenAI | 94.3% | 76.0 | $10.00 | |
| 10 | Gemini 3.6 FlashGoogle | 94.1% | 62.5 | $3.75 | |
| 11 | Grok 4.6xAI | 94.0% | 65.7 | $6.00 | |
| 12 | Claude Opus 5Anthropic | 93.9% | 83.0 | $25.00 | |
| 13 | GPT-5.6 SolOpenAI | 93.5% | 73.5 | $20.00 | |
| 14 | Grok 4.5xAI | 93.4% | 58.6 | $6.00 | |
| 15 | GPT-5.6 TerraOpenAI | 93.3% | 63.9 | $12.00 | |
| 16 | GPT-5.4OpenAI | 93.3% | 64.4 | $15.00 | |
| 17 | DeepSeek V4.1 Flash 1017DeepSeek | 93.2% | 70.3 | $0.60 | |
| 18 | Kimi K3Moonshot AI | 93.1% | 62.3 | $15.00 | |
| 19 | Gemini 3.5 FlashGoogle | 92.8% | 39.8 | $9.00 | |
| 20 | Grok 4.7xAI | 92.7% | 66.7 | $6.00 | |
| 21 | Qwen3.8 Max (0902)Qwen | 92.7% | 62.4 | $6.00 | |
| 22 | GLM 5.2Z.ai | 91.9% | 61.1 | $4.40 | |
| 23 | DeepSeek V4 Pro 0813DeepSeek | 91.7% | 56.1 | $1.98 | |
| 24 | GPT-5.6 LunaOpenAI | 91.6% | 58.0 | $1.20 | |
| 25 | GPT-5.2OpenAI | 91.4% | 55.9 | $14.00 | |
| 26 | Claude Opus 4.8Anthropic | 91.0% | 65.0 | $25.00 | |
| 27 | DeepSeek V4 Flash 0731DeepSeek | 91.0% | 51.8 | $0.396 | |
| 28 | DeepSeek V4 Pro 0423DeepSeek | 90.9% | 38.7 | $2.976 | |
| 29 | MiniMax M3MiniMax | 90.9% | 39.2 | $1.20 | |
| 30 | Qwen3.7 MaxQwen | 90.9% | 54.1 | $4.425 | |
| 31 | GLM 5.3Z.ai | 90.9% | 70.6 | $4.40 | |
| 32 | Kimi K2.6Moonshot AI | 90.8% | 52.1 | $4.00 | |
| 33 | GPT-5.5OpenAI | 90.7% | 71.3 | $30.00 | |
| 34 | Claude Opus 5.5Anthropic | 90.6% | 88.8 | $20.00 | |
| 35 | Claude Opus 4.6Anthropic | 90.5% | 57.2 | $25.00 | |
| 36 | Claude Sonnet 5Anthropic | 90.5% | 57.7 | $10.00 | |
| 37 | GPT-6 LunaOpenAI | 90.5% | 57.3 | $0.50 | |
| 38 | Claude Opus 4.7Anthropic | 90.2% | 67.5 | $25.00 | |
| 39 | GLM 5.3 FlashZ.ai | 90.2% | 57.7 | $0.50 | |
| 40 | GLM 5.1Z.ai | 89.9% | 49.2 | $4.40 | |
| 41 | DeepSeek V4.1 FlashDeepSeek | 89.8% | 62.5 | $0.84 | |
| 42 | Claude Haiku 5.5Anthropic | 89.6% | 70.7 | $0.50 | |
| 43 | Gemini 3 Flash PreviewGoogle | 89.4% | 43.7 | $3.00 | |
| 44 | Grok 4.3xAI | 88.8% | 37.4 | $2.50 | |
| 45 | T Inkling SmallThinking Machines | 88.5% | 27.3 | $1.20 | |
| 46 | Qwen3.6 PlusQwen | 88.4% | 44.1 | $1.95 | |
| 47 | T InklingThinking Machines | 88.3% | 31.5 | $4.05 | |
| 48 | Mistral Large 4Mistral AI | 88.1% | 56.4 | $4.18 | |
| 49 | Kimi K2.7 CodeMoonshot AI | 87.9% | 53.4 | $4.00 | |
| 50 | Qwen3.7 PlusQwen | 87.9% | 33.2 | $1.28 | |
| 51 | GLM 5Z.ai | 87.8% | 41.7 | $3.20 | |
| 52 | GPT-5.1OpenAI | 87.6% | 40.4 | $10.00 | |
| 53 | Claude Sonnet 4.6Anthropic | 87.4% | 57.3 | $15.00 | |
| 54 | GPT-5.4 MiniOpenAI | 86.9% | 35.8 | $4.50 | |
| 55 | Qwen3.5 397B A17BQwen | 86.4% | Not benchmarked | $2.34 | |
| 56 | GPT-5OpenAI | 86.2% | 38.8 | $10.00 | |
| 57 | Claude Opus 4.5Anthropic | 86.0% | 44.0 | $25.00 | |
| 58 | Claude Fable 5Anthropic | 85.9% | 79.3 | $50.00 | |
| 59 | Qwen3.6 27BQwen | 85.9% | 36.8 | $2.70 | |
| 60 | Nemotron 3 UltraNVIDIA | 85.4% | 30.0 | $2.40 | |
| 61 | Gemini 2.5 ProGoogle | 85.3% | 19.1 | $10.00 | |
| 62 | Gemini 2.5 Pro Preview 06-05Google | 84.8% | Not benchmarked | $10.00 | |
| 63 | Qwen3.5 Plus 2026-04-20Qwen | 84.8% | 39.5 | $1.80 | |
| 64 | Qwen3.6 35B A3BQwen | 84.8% | Not benchmarked | $1.00 | |
| 65 | Qwen3.5-35B-A3BQwen | 83.5% | Not benchmarked | $1.30 | |
| 66 | Gemini 3.5 Flash LiteGoogle | 83.3% | 49.0 | $2.50 | |
| 67 | Qwen3.6 FlashQwen | 83.3% | Not benchmarked | $1.125 | |
| 68 | GLM 4.7Z.ai | 83.3% | 55.1 | $2.20 | |
| 69 | Claude Sonnet 4.5Anthropic | 82.3% | 39.5 | $15.00 | |
| 70 | Qwen3.5-FlashQwen | 82.3% | 41.1 | $0.26 | |
| 71 | Qwen3.7 FlashQwen | 82.3% | Not benchmarked | $0.13 | |
| 72 | Gemini 3.1 Flash LiteGoogle | 81.8% | Not benchmarked | $1.50 | |
| 73 | o3OpenAI | 81.8% | Not benchmarked | $8.00 | |
| 74 | Qwen3 235B A22B Thinking 2507Qwen | 80.1% | Not benchmarked | $3.50 | |
| 75 | Claude 3.7 SonnetRetiredAnthropic | 79.7% | Not benchmarked | $15.00 | |
| 76 | o4 MiniOpenAI | 79.6% | Not benchmarked | $4.40 | |
| 77 | Qwen3.5-9BQwen | 79.0% | Not benchmarked | $0.15 | |
| 78 | GPT-5.4 NanoOpenAI | 78.5% | 44.8 | $1.25 | |
| 79 | Claude Opus 4.1Anthropic | 77.3% | Not benchmarked | $75.00 | |
| 80 | o3 MiniOpenAI | 77.0% | Not benchmarked | $4.40 | |
| 81 | o1OpenAI | 76.8% | Not benchmarked | $60.00 | |
| 82 | R1 0528DeepSeek | 76.3% | Not benchmarked | $2.18 | |
| 83 | Claude Opus 4RetiredAnthropic | 76.3% | Not benchmarked | $75.00 | |
| 84 | Gemma 4 31BGoogle | 75.8% | 39.0 | $0.40 | |
| 85 | gpt-oss-120bOpenAI | 75.8% | Not benchmarked | $0.50 | |
| 86 | Grok 3RetiredxAI | 75.8% | Not benchmarked | $15.00 | |
| 87 | GPT-5 MiniOpenAI | 75.0% | 34.2 | $2.00 | |
| 88 | Gemma 4 26B A4BGoogle | 73.2% | Not benchmarked | $0.60 | |
| 89 | R1RetiredDeepSeek | 71.7% | Not benchmarked | $2.50 | |
| 90 | Claude Haiku 4.5Anthropic | 71.2% | 35.6 | $5.00 | |
| 91 | DeepSeek V3DeepSeek | 71.2% | Not benchmarked | $0.9594 | |
| 92 | GPT-5 NanoOpenAI | 69.4% | Not benchmarked | $0.40 | |
| 93 | DeepSeek V3 0324DeepSeek | 67.6% | Not benchmarked | $1.07 | |
| 94 | GPT-4.1OpenAI | 66.9% | Not benchmarked | $8.00 | |
| 95 | GPT-4.1 MiniOpenAI | 65.8% | Not benchmarked | $1.60 | |
| 96 | Qwen3 32BQwen | 65.7% | Not benchmarked | $0.40 | |
| 97 | Gemini 2.0 FlashRetiredGoogle | 64.1% | Not benchmarked | $0.40 | |
| 98 | Qwen3 14BQwen | 63.8% | Not benchmarked | $0.23 | |
| 99 | o1-miniRetiredOpenAI | 62.4% | Not benchmarked | $4.40 | |
| 100 | Qwen3 30B A3BQwen | 61.7% | Not benchmarked | $0.50 | |
| 101 | gpt-oss-20bOpenAI | 60.8% | Not benchmarked | $0.15 | |
| 102 | GLM 4.7 FlashZ.ai | 60.5% | Not benchmarked | $0.40 | |
| 103 | Gemini 1.5 ProRetiredGoogle | 57.2% | Not benchmarked | $5.00 | |
| 104 | Phi 4Microsoft | 56.1% | Not benchmarked | $0.14 | |
| 105 | Qwen3 30B A3B Instruct 2507Qwen | 55.6% | Not benchmarked | $0.30 | |
| 106 | Claude 3.5 SonnetRetiredAnthropic | 55.3% | Not benchmarked | $15.00 | |
| 107 | Grok 2RetiredxAI | 53.8% | Not benchmarked | $10.00 | |
| 108 | Llama 4 ScoutMeta Llama | 51.8% | Not benchmarked | $0.50 | |
| 109 | GPT-4o (2024-08-06)OpenAI | 49.2% | Not benchmarked | $10.00 | |
| 110 | Qwen2.5 72B InstructQwen | 49.1% | Not benchmarked | $0.40 | |
| 111 | Mistral Large 2407Mistral AI | 49.0% | Not benchmarked | $6.00 | |
| 112 | GPT-4.1 NanoOpenAI | 48.9% | Not benchmarked | $0.40 | |
| 113 | GPT-4o (2024-05-13)OpenAI | 48.9% | Not benchmarked | $15.00 | |
| 114 | Qwen-PlusQwen | 48.1% | Not benchmarked | $0.78 | |
| 115 | GPT-4o (2024-11-20)OpenAI | 47.9% | Not benchmarked | $10.00 | |
| 116 | Gemma 3 27BGoogle | 47.7% | Not benchmarked | $0.30 | |
| 117 | Llama 3.3 70B InstructMeta Llama | 47.4% | Not benchmarked | $0.72 | |
| 118 | Gemini 1.5 FlashRetiredGoogle | 47.3% | Not benchmarked | $0.30 | |
| 119 | Claude 3 OpusRetiredAnthropic | 47.2% | Not benchmarked | $75.00 | |
| 120 | Llama 3.1 70B InstructMeta Llama | 44.2% | Not benchmarked | $0.56 | |
| 121 | WizardLM-2 8x22BRetiredMicrosoft | 43.4% | Not benchmarked | $0.62 | |
| 122 | Claude 3 SonnetRetiredAnthropic | 40.6% | Not benchmarked | $15.00 | |
| 123 | Gemma 3 12BGoogle | 39.5% | Not benchmarked | $0.1625 | |
| 124 | Claude 3.5 HaikuRetiredAnthropic | 38.1% | Not benchmarked | $4.00 | |
| 125 | GPT-4o-mini (2024-07-18)OpenAI | 37.7% | Not benchmarked | $0.60 | |
| 126 | Gemma 2 27BGoogle | 36.5% | Not benchmarked | $0.65 | |
| 127 | Claude 3 HaikuRetiredAnthropic | 36.3% | Not benchmarked | $1.25 | |
| 128 | Qwen2.5 7B InstructQwen | 35.5% | Not benchmarked | $0.20 | |
| 129 | Claude 2RetiredAnthropic | 34.7% | Not benchmarked | $24.00 | |
| 130 | Granite 4.0 MicroIBM Granite | 28.3% | Not benchmarked | $0.112 | |
| 131 | Llama 3.1 8B InstructMeta Llama | 27.0% | Not benchmarked | $0.08 | |
| 132 | Llama 3.2 1B InstructMeta Llama | 23.9% | Not benchmarked | $0.201 | |
| 133 | Gemma 3 4BGoogle | 23.2% | Not benchmarked | $0.075 |
Only models in our catalog are listed, and each appears once with its best published run. 215 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.