ProofBench v1.1
Writing rigorous mathematical proofs, graded for correctness. Vals AI's own benchmark; Epoch copies earlier versions, so they share weight. Run by Vals AI.
| Category | Math |
|---|---|
| Metric | Accuracy (%) |
| Source | Vals AI (evaluator) |
| License | None stated |
| Benchmark page | www.vals.ai |
| Status | Scored |
| Weight | 1.0 within Math |
| Fit | Discrimination 4.20, noise 0.78 |
| Lineage | Shares weight with proofbench |
Results
| # | Model | Provider | Result | Date |
|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 Max | Anthropic | 100.0% | 2026-11-09 |
| 2 | Claude Opus 5.5 Max | Anthropic | 100.0% | 2026-11-09 |
| 3 | Claude Fable 5.1 Max | Anthropic | 100.0% | 2026-11-09 |
| 4 | Claude Fable 5.2 High | Anthropic | 100.0% | 2026-11-09 |
| 5 | Gemini 4 Argon High | 99.0% | 2026-11-09 | |
| 6 | GPT-6.1 Sol Max | OpenAI | 99.0% | 2026-11-09 |
| 7 | GPT-6 Astra Max | OpenAI | 99.0% | 2026-11-09 |
| 8 | Claude Opus 5 Max | Anthropic | 99.0% | 2026-11-09 |
| 9 | Claude Fable 5 Max | Anthropic | 95.0% | 2026-11-09 |
| 10 | Muse Wonder Full Max | Meta | 94.0% | 2026-11-09 |
| 11 | Kimi K3.1 Pro Max | Moonshot AI | 91.0% | 2026-11-09 |
| 12 | Claude Haiku 5.5 Max | Anthropic | 89.0% | 2026-11-09 |
| 13 | Kimi K3 Max | Moonshot AI | 87.0% | 2026-11-09 |
| 14 | Muse Wonder Max | Meta | 84.0% | 2026-11-09 |
| 15 | GPT-6 Sol Max | OpenAI | 83.0% | 2026-11-09 |
| 16 | GPT-5.6 Sol Max | OpenAI | 83.0% | 2026-11-09 |
| 17 | Claude Sonnet 5 Max | Anthropic | 77.0% | 2026-11-09 |
| 18 | Hy4 preview | Tencent | 75.0% | 2026-11-09 |
| 19 | GPT-5.6 Terra Max | OpenAI | 74.0% | 2026-11-09 |
| 20 | MiMo-V2.6-Pro | Xiaomi | 70.0% | 2026-11-09 |
| 21 | GPT-6 Luna Max | OpenAI | 64.0% | 2026-11-09 |
| 22 | MiMo-V2.6-Flash | Xiaomi | 63.0% | 2026-11-09 |
| 23 | Ember-1 | Fireworks | 61.0% | 2026-11-09 |
| 24 | GPT-5.6 Luna Max | OpenAI | 60.0% | 2026-11-09 |
| 25 | Qwen3.8 Max | Alibaba | 58.0% | 2026-11-09 |
| 26 | Gemini 3.7 Flash High | 58.0% | 2026-11-09 | |
| 27 | Muse Spark 1.3 Max | Meta | 58.0% | 2026-11-09 |
| 28 | DeepSeek V4 Flash 0731 High | DeepSeek | 56.0% | 2026-11-09 |
| 29 | Muse Spark 1.3 Extra High | Meta | 55.0% | 2026-11-09 |
| 30 | DeepSeek V4.1 Flash High | DeepSeek | 54.0% | 2026-11-09 |
| 31 | Grok 4.6 High | xAI | 51.0% | 2026-11-09 |
| 32 | DeepSeek V4 Pro 0813 Max | DeepSeek | 50.0% | 2026-11-09 |
| 33 | GLM-5.3 Max | Zhipu AI | 49.0% | 2026-11-09 |
| 34 | Gemini 3.8 Flash High | 48.0% | 2026-11-09 | |
| 35 | Muse Spark 1.2 Extra High | Meta | 43.0% | 2026-11-09 |
| 36 | Step 5 Preview | StepFun | 42.0% | 2026-11-09 |
| 37 | Grok 4.5 High | xAI | 31.0% | 2026-11-09 |
| 38 | Gemini 3.5 Flash High | 31.0% | 2026-11-09 | |
| 39 | Grok 4.7 Extra High | xAI | 26.0% | 2026-11-09 |
| 40 | Gemini 3.1 Pro Preview High | 26.0% | 2026-11-09 | |
| 41 | MiMo-V2.5-Pro | Xiaomi | 22.0% | 2026-11-09 |
| 42 | GLM-5.3-Flash Max | Zhipu AI | 21.0% | 2026-11-09 |
| 43 | MiniMax-M3 | MiniMax | 18.0% | 2026-11-09 |
| 44 | Beam 1.0 | Reflection | 18.0% | 2026-11-09 |
| 45 | MiMo-V2.5 | Xiaomi | 16.0% | 2026-11-09 |
| 46 | Qwen3.8 27B Extra High | Alibaba | 16.0% | 2026-11-09 |
| 47 | DeepSeek V4 Pro Max | DeepSeek | 16.0% | 2026-11-09 |
| 48 | Mistral Large 4 High | Mistral | 10.0% | 2026-11-09 |
| 49 | Mistral Medium 3.5 High | Mistral | 9.0% | 2026-11-09 |
| 50 | Inkling Small | Thinking Machines | 6.0% | 2026-11-09 |
| 51 | Mercury 2.5 High | Inception | 3.0% | 2026-11-09 |
| 52 | Inkling | Thinking Machines | 0.0% | 2026-11-09 |