Proofbench — Epoch AI
Proofbench — Epoch AI is run by Epoch AI. It has ranked 84 models, of which the catalogue holds 78, scoring from 0 to 100.
measured by Epoch AI · the board itself
| Place | Model | Metric | Score |
|---|---|---|---|
| 1stof 84 | Claude Fable 5.1 | Accuracy (max) | 100 |
| 2ndof 84 | Claude Opus 5.5 | Accuracy (max) | 100 |
| 3rdof 84 | Claude Sonnet 5.5 | Accuracy (max) | 100 |
| 4thof 84 | Claude Opus 5 | Accuracy (max) | 99 |
| 5thof 84 | GPT 6 Astra | Accuracy | 99 |
| 6thof 84 | GPT 6.1 Sol | Accuracy | 99 |
| 7thof 84 | Gemini 4 Argon | Accuracy | 99 |
| 8thof 84 | Claude Fable 5 | Accuracy (max) | 95 |
| 9thof 84 | Kimi K3.1 Pro | Accuracy | 91 |
| 10thof 84 | Claude Haiku 5.5 | Accuracy (max) | 89 |
| 11thof 84 | Kimi K3 | Accuracy | 87 |
| 13thof 84 | GPT-5.6 Sol | Accuracy (max) | 83 |
| 14thof 84 | GPT 6 Sol | Accuracy | 83 |
| 15thof 84 | Claude Sonnet 5 | Accuracy (max) | 77 |
| 16thof 84 | Hy4-preview | Accuracy | 75 |
| 17thof 84 | GPT-5.6 Terra | Accuracy (xhigh) | 74 |
| 18thof 84 | MiMo V2.6 Pro | Accuracy | 70 |
| 19thof 84 | Claude Opus 4.8 | Accuracy (max) | 69 |
| 20thof 84 | GPT 6 Luna | Accuracy | 64 |
| 21stof 84 | MiMo V2.6 Flash | Accuracy | 63 |
| 22ndof 84 | GPT-5.6 Luna | Accuracy (max) | 60 |
| 23rdof 84 | Gemini 3.7 Flash | Accuracy | 58 |
| 24thof 84 | Qwen 3.8 Max | Accuracy | 58 |
| 25thof 84 | Muse Spark 1.3 | Accuracy (max) | 58 |
| 26thof 84 | GPT-5.4 | Accuracy (xhigh) | 56 |
| 27thof 84 | DeepSeek V4 Flash (0731) | Accuracy | 56 |
| 29thof 84 | Claude Opus 4.7 | Accuracy (max) | 54 |
| 30thof 84 | DeepSeek V4.1 Flash | Accuracy | 54 |
| 31stof 84 | Grok 4.6 | Accuracy | 51 |
| 32ndof 84 | GPT-5.5 | Accuracy (xhigh) | 50 |
| 33rdof 84 | Claude Opus 4.6 | Accuracy (max) | 50 |
| 34thof 84 | DeepSeek V4 Pro 0813 | Accuracy | 50 |
| 35thof 84 | GLM-5.3 | Accuracy (max) | 49 |
| 36thof 84 | Gemini 3.8 Flash | Accuracy | 48 |
| 37thof 84 | Claude Sonnet 4.6 | Accuracy (max) | 45 |
| 38thof 84 | Muse Spark 1.2 | Accuracy | 43 |
| 39thof 84 | Step 5 | Accuracy | 42 |
| 40thof 84 | Muse Spark 1.1 | Accuracy | 39 |
| 41stof 84 | Claude Opus 4.5 | Accuracy | 36 |
| 42ndof 84 | Gemini 3.6 Flash | Accuracy | 36 |
| 43rdof 84 | GLM 5.2 | Accuracy (max) | 35 |
| 44thof 84 | Grok 4.5 | Accuracy (high) | 31 |
| 45thof 84 | Gemini 3.5 Flash | Accuracy (high) | 31 |
| 46thof 84 | Qwen3.7 Max | Accuracy (max) | 26 |
| 47thof 84 | Gemini 3.1 Pro Preview | Accuracy | 26 |
| 48thof 84 | Grok 4.7 | Accuracy | 26 |
| 49thof 84 | GLM 5.1 | Accuracy | 22.222 |
| 50thof 84 | MiMo-V2.5-Pro | Accuracy | 22 |
| 51stof 84 | GPT-5.4 mini | Accuracy (xhigh) | 21 |
| 52ndof 84 | GLM 5.3 Flash | Accuracy (max) | 21 |
| 53rdof 84 | gemini-3-pro-preview | Accuracy | 20 |
| 54thof 84 | Claude Sonnet 4.5 | Accuracy | 19 |
| 55thof 84 | MiniMax M3 | Accuracy | 18 |
| 56thof 84 | GPT-5 | Accuracy (high) | 18 |
| 59thof 84 | Kimi K2.6 | Accuracy | 16 |
| 60thof 84 | MiMo-V2.5 | Accuracy | 16 |
| 61stof 84 | DeepSeek V4 Pro | Accuracy (max) | 16 |
| 62ndof 84 | Qwen3.8 27B | Accuracy | 16 |
| 63rdof 84 | Gemini 3 Flash Preview | Accuracy | 15 |
| 64thof 84 | GPT-5.2 | Accuracy (xhigh) | 15 |
| 65thof 84 | Grok 4.20 | Accuracy | 14 |
| 66thof 84 | Gemini 3.5 Flash-Lite | Accuracy | 13 |
| 67thof 84 | GPT-5 nano | Accuracy (high) | 12 |
| 68thof 84 | Grok 4.3 | Accuracy (high) | 11 |
| 70thof 84 | GPT-5 mini | Accuracy (high) | 9 |
| 71stof 84 | GPT-5.1 Codex Max | Accuracy (max) | 9 |
| 73rdof 84 | DeepSeek v3.2 | Accuracy | 8 |
| 74thof 84 | GLM-4.7 | Accuracy | 6 |
| 75thof 84 | Inkling Small | Accuracy | 6 |
| 76thof 84 | GPT-5.4 nano | Accuracy (high) | 5 |
| 77thof 84 | Grok 4.1 Fast Reasoning | Accuracy | 4 |
| 78thof 84 | MiniMax M2.5 | Accuracy | 4 |
| 79thof 84 | MiniMax M2.7 | Accuracy | 3 |
| 80thof 84 | Mercury 2.5 | Accuracy | 3 |
| 81stof 84 | Nemotron 3 Ultra | Accuracy | 2 |
| 82ndof 84 | Laguna-XS.2 | Accuracy | 0 |
| 83rdof 84 | Laguna M.1 | Accuracy | 0 |
| 84thof 84 | Inkling | Accuracy | 0 |
Read from the board on 2026-10-31