MathArena - IMProofBench Proofs: leaderboard
Tests writing complete, rigorous mathematical proofs for 39 research-level problems spanning algebraic geometry, combinatorics, graph theory, and stochastic analysis, graded by expert mathematicians.
Metric: Accuracy (%). Source: matharena.ai. Saturation forecast: Around March 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (High) | 49.56 | #98 (GPT-5) |
| 2 | Gemini 2.5 Pro | 34.2 | #152 |
| 3 | O4 Mini (High) | 31.37 | #180 (O4 Mini) |
| 4 | Grok 4 | 28.98 | #181 |
| 5 | Claude Opus 4.1 (Thinking) | 24.62 | #140 (Claude Opus 4.1) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=matharena-improofbench-proofs · How It Works · Data refreshed daily, snapshot 2026-10-20.