GPQA diamond
Each model’s result as Epoch AI’s Benchmarking Hub publishes it, run by Epoch or collected from the source it cites. Facthory has not run them.
Data: Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, licensed CC BY 4.0. Read on 16 Nov 2026.
The chart shows the 25 best and the table the 100 best of 339 results.
| # | Model | Organization | Result | Evidence |
|---|---|---|---|---|
| 1 | Gemini 4 Carbon (high) API access | Google DeepMind | 96.5% ± 1.2% | – |
| 2 | Gemini 4 Argon (high) API access | Google DeepMind | 96.4% ± 1.2% | Log or source |
| 3 | Claude Fable 5.2 (with fallback, max) API access | Anthropic | 96% ± 1.3% | – |
| 4 | GPT-6 Astra (max) API access | OpenAI | 95.8% ± 1.4% | Log or source |
| 5 | Claude Sonnet 5.5 (max) API access | Anthropic | 95.6% ± 1.4% | Log or source |
| 6 | Gemini 3.8 Flash (high) API access | Google DeepMind | 95.4% ± 1.4% | Log or source |
| 7 | GPT-6.1 Sol (max) API access | OpenAI | 95.4% ± 1.4% | Log or source |
| 8 | Gemini 3.7 Flash (high) API access | Google DeepMind | 94.8% ± 1.3% | Log or source |
| 9 | GPT-5.4 Pro (xhigh) API access | OpenAI | 94.6% ± 1.6% | – |
| 10 | Gemini 3.1 Pro Preview API access | Google DeepMind | 94.4% ± 1.6% | Log or source |
| 11 | GPT-6 Sol (max) API access | OpenAI | 94.3% ± 1.5% | Log or source |
| 12 | GPT-6.1 Terra (max) API access | OpenAI | 94.3% ± 1.5% | Log or source |
| 13 | Kimi K3.1 Pro (Together, max) Open weights (restricted use) | Moonshot | 94.2% ± 1.4% | Log or source |
| 14 | Gemini 3.6 Flash API access | Google DeepMind | 94.1% ± 1.4% | Log or source |
| 15 | Gemini 3.1 Pro Preview API access | Google DeepMind | 94.1% ± 1.7% | Log or source |
| 16 | Claude Opus 5.5 (with fallback, max) API access | Anthropic | 94.1% ± 1.5% | – |
| 17 | GPT-5.5 (xhigh) API access | OpenAI | 94% ± 1.5% | Log or source |
| 18 | Grok 4.6 (high) API access | xAI | 94% ± 1.4% | Log or source |
| 19 | GPT-5.5 Pro (xhigh) API access | OpenAI | 93.9% ± 1.6% | Log or source |
| 20 | Claude Fable 5.2 (max) API access | Anthropic | 93.9% ± 1.6% | – |
| 21 | Claude Fable 5 (with fallback, max) API access | Anthropic | 93.9% ± 1.7% | – |
| 22 | Claude Opus 5 (max) API access | Anthropic | 93.9% ± 1.5% | Log or source |
| 23 | GPT-5.6 Sol (max) API access | OpenAI | 93.5% ± 1.6% | Log or source |
| 24 | Grok 4.5 (high) API access | xAI | 93.4% ± 1.4% | Log or source |
| 25 | GPT-5.6 Terra (max) API access | OpenAI | 93.3% ± 1.5% | Log or source |
| 26 | GPT-5.4 (xhigh) API access | OpenAI | 93.3% ± 1.8% | Log or source |
| 27 | DeepSeek V4.1 Flash 1017 (max) Open weights (unrestricted) | DeepSeek | 93.2% ± 1.6% | Log or source |
| 28 | Grok 4.6 (xhigh) API access | xAI | 93.2% ± 1.5% | Log or source |
| 29 | Doubao Seed Evolving (Nov 2026, high) API access | ByteDance | 93.1% ± 1.6% | Log or source |
| 30 | Kimi K3 (max) Open weights (non-commercial) | Moonshot | 93.1% ± 1.5% | Log or source |
| 31 | Claude Opus 5 API access | Anthropic | 92.9% ± 1.8% | Log or source |
| 32 | Gemini 3.5 Flash API access | Google DeepMind | 92.8% ± 1.6% | Log or source |
| 33 | Grok 4.7 (xhigh) API access | xAI | 92.7% ± 1.6% | Log or source |
| 34 | Qwen3.8 Max (xhigh) API access | Alibaba | 92.7% ± 1.7% | Log or source |
| 35 | Gemini 3 Pro Preview API access | Google DeepMind | 92.6% ± 1.7% | Log or source |
| 36 | Qwen3.8 Max (0902) (xhigh) API access | Alibaba | 92.3% ± 1.7% | Log or source |
| 37 | Kimi K3 (high) Open weights (non-commercial) | Moonshot | 91.9% ± 1.9% | Log or source |
| 38 | GLM-5.2 Open weights (unrestricted) | Z.ai (Zhipu AI) | 91.9% ± 1.6% | Log or source |
| 39 | DeepSeek V4 Pro 0813 (max) Open weights (unrestricted) | DeepSeek | 91.7% ± 1.5% | Log or source |
| 40 | GPT-5.6 Luna (max) API access | OpenAI | 91.6% ± 1.7% | Log or source |
| 41 | GPT-5.2 (xhigh) API access | OpenAI | 91.4% ± 1.8% | Log or source |
| 42 | Claude Opus 4.8 API access | Anthropic | 91% ± 1.9% | Log or source |
| 43 | DeepSeek V4 Flash 0731 (max) Open weights (unrestricted) | DeepSeek | 91% ± 1.8% | Log or source |
| 44 | MiniMax-M3 Open weights (restricted use) | MiniMax | 90.9% ± 2% | Log or source |
| 45 | DeepSeek v4 Pro (high) Open weights (unrestricted) | DeepSeek | 90.9% ± 2% | Log or source |
| 46 | GLM-5.3 (max) Open weights (restricted use) | Z.ai (Zhipu AI) | 90.9% ± 1.6% | Log or source |
| 47 | Qwen3.7 Max API access | Alibaba | 90.9% ± 2% | Log or source |
| 48 | Kimi K2.6 Open weights (unrestricted) | Moonshot | 90.8% ± 1.7% | Log or source |
| 49 | GPT-5.5 (low) API access | OpenAI | 90.7% ± 1.8% | Log or source |
| 50 | Claude Opus 5.5 (max) API access | Anthropic | 90.6% ± 1.9% | Log or source |
| 51 | Claude Opus 4.6 (32k thinking) API access | Anthropic | 90.5% ± 1.7% | Log or source |
| 52 | Claude Sonnet 5 (xhigh) API access | Anthropic | 90.5% ± 1.8% | Log or source |
| 53 | GPT-6 Luna (max) API access | OpenAI | 90.5% ± 1.9% | Log or source |
| 54 | Claude Opus 4.7 (xhigh) API access | Anthropic | 90.2% ± 1.8% | Log or source |
| 55 | GLM-5.3-Flash (max) Open weights (unrestricted) | Z.ai (Zhipu AI) | 90.2% ± 1.7% | Log or source |
| 56 | Mistral Large 4 (high) Open weights (restricted use) | Mistral AI | 90.2% ± 1.8% | Log or source |
| 57 | GLM-5.1 Open weights (unrestricted) | Z.ai (Zhipu AI) | 89.9% ± 2.1% | Log or source |
| 58 | GPT-5.4 (high) API access | OpenAI | 89.9% ± 2.1% | Log or source |
| 59 | GPT-5.6 Sol (low) API access | OpenAI | 89.9% ± 2.1% | Log or source |
| 60 | DeepSeek V4.1 Flash (max) Open weights (unrestricted) | DeepSeek | 89.8% ± 1.7% | Log or source |
| 61 | Muse Spark API access | Meta AI | 89.8% ± 2.2% | – |
| 62 | Claude Haiku 5.5 (max) API access | Anthropic | 89.6% ± 2% | Log or source |
| 63 | DeepSeek v4 Pro (max) Open weights (unrestricted) | DeepSeek | 89.6% ± 1.7% | Log or source |
| 64 | Doubao Seed 2.1 Turbo (high) API access | ByteDance | 89.6% ± 1.8% | Log or source |
| 65 | Gemini 3 Flash Preview API access | Google DeepMind | 89.4% ± 2.2% | Log or source |
| 66 | Doubao Seed 2.1 Pro (high) API access | ByteDance | 89.3% ± 1.8% | Log or source |
| 67 | grok-4.20-0309-reasoning API access | xAI | 89.3% ± 1.9% | Log or source |
| 68 | Gemini 3.5 Flash API access | Google DeepMind | 88.9% ± 2.2% | Log or source |
| 69 | GPT-5.4 (medium) API access | OpenAI | 88.9% ± 2.2% | Log or source |
| 70 | grok-4.3_high Hosted access (no API) | xAI | 88.8% ± 2% | Log or source |
| 71 | Claude Opus 4.6 (64k thinking) API access | Anthropic | 88.8% ± 1.9% | Log or source |
| 72 | Muse Spark Open (xhigh) Open weights (restricted use) | Meta AI | 88.6% ± 1.9% | Log or source |
| 73 | Inkling Small (xhigh) Open weights (unrestricted) | Thinking Machines | 88.5% ± 1.9% | Log or source |
| 74 | Claude Opus 4.6 (max) API access | Anthropic | 88.4% ± 2.3% | Log or source |
| 75 | Claude Opus 4.8 API access | Anthropic | 88.4% ± 2.3% | Log or source |
| 76 | Qwen 3.6 Plus (2026-04-02) API access | Alibaba | 88.4% ± 2.3% | Log or source |
| 77 | Inkling (xhigh) Open weights (unrestricted) | Thinking Machines | 88.3% ± 1.9% | Log or source |
| 78 | Doubao-Seed-2.0 Pro (high) API access | ByteDance | 88.2% ± 1.9% | Log or source |
| 79 | GPT-5.2 (high) API access | OpenAI | 88.2% ± 1.9% | Log or source |
| 80 | Doubao Seed 2.0 Lite (Apr 2026, high) API access | ByteDance | 88.1% ± 1.9% | Log or source |
| 81 | Mistral Large 4 Preview (high) Open weights (restricted use) | Mistral AI | 88.1% ± 1.9% | Log or source |
| 82 | Claude Opus 5 (low) API access | Anthropic | 87.9% ± 2.3% | Log or source |
| 83 | GLM-5.2 Open weights (unrestricted) | Z.ai (Zhipu AI) | 87.9% ± 2.3% | Log or source |
| 84 | GPT-5.2 (medium) API access | OpenAI | 87.9% ± 1.9% | Log or source |
| 85 | Kimi K2.7 Code Open weights (unrestricted) | Moonshot | 87.9% ± 2.3% | Log or source |
| 86 | Qwen3.7 Plus API access | Alibaba | 87.9% ± 2.3% | Log or source |
| 87 | GLM-5 Open weights (unrestricted) | Z.ai (Zhipu AI) | 87.8% ± 2.3% | Log or source |
| 88 | GPT-5.1 (high) API access | OpenAI | 87.6% ± 1.9% | Log or source |
| 89 | Kimi K2.5 (Fireworks) Open weights (unrestricted) | Moonshot | 87.6% ± 1.9% | Log or source |
| 90 | Claude Sonnet 4.6 (32k thinking) API access | Anthropic | 87.4% ± 2% | Log or source |
| 91 | GPT-5.6 Terra (low) API access | OpenAI | 87.4% ± 2.4% | Log or source |
| 92 | Qwen 3.6 Max (Preview) | Alibaba | 87.4% ± 2.4% | Log or source |
| 93 | grok-4-0709 API access | xAI | 87% ± 2% | – |
| 94 | GPT-5.4 mini (xhigh) API access | OpenAI | 86.9% ± 2.4% | Log or source |
| 95 | Claude Opus 4.7 (max) API access | Anthropic | 86.4% ± 2.4% | Log or source |
| 96 | Gemini 3.5 Flash API access | Google DeepMind | 86.4% ± 2.4% | Log or source |
| 97 | Gemini 3.6 Flash API access | Google DeepMind | 86.4% ± 2.4% | Log or source |
| 98 | Qwen3.5 397B-A17B Open weights (unrestricted) | Alibaba | 86.4% ± 2.4% | Log or source |
| 99 | GPT-5 (high) API access | OpenAI | 86.2% ± 2.1% | Log or source |
| 100 | Claude Opus 4.5 (32k thinking) API access | Anthropic | 86% ± 2.1% | Log or source |
