MathArena - IMProofBench Final Answers — leaderboard
Metric: Accuracy (%). Source: matharena.ai. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 83.52 |
| 2 | Claude Opus 4.6 (High) | 80.74 |
| 3 | GPT-5.2 (xHigh) | 73.17 |
| 4 | Gemini 3 Pro (Preview) | 72.28 |
| 5 | GPT-5.1 (High) | 62.01 |
| 6 | GPT-5 Pro | 59.29 |
| 7 | Grok 4 | 57.48 |
| 8 | GPT-5 (High) | 53.8 |
| 9 | O4 Mini (High) | 45.22 |
| 10 | Grok 4 Fast (Reasoning) | 44.05 |
| 11 | O3 (High) | 41.27 |
| 12 | Gemini 2.5 Pro | 39.15 |
| 13 | Claude Opus 4.1 (Thinking) | 38.42 |
| 14 | Gemini 2.5 Flash (Thinking) | 37.46 |
| 15 | GPT-OSS-120B (High) | 37.32 |
Interactive version: theaggregate.ai/benchmark?slug=matharena-improofbench-final-answers · How the rankings work · Data refreshed daily, snapshot 2026-07-22.