MathArena - ARXIVLEAN March: leaderboard
Lean-formalized research-level math problems from March arXiv papers, requiring machine-verifiable formal proofs rather than natural-language solutions.
Metric: Accuracy (%). Source: matharena.ai. Saturation forecast: Around July 2028. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 17.07 | #31 (GPT-5.5) |
| 2 | GPT-5.4 (xHigh) | 17.07 | #79 (GPT-5.4) |
| 3 | Gemini 3.1 Pro (Preview) | 14.63 | #58 |
| 4 | Claude Opus 4.7 (High) | 14.63 | #49 (Claude Opus 4.7) |
| 5 | GLM-5.1 | 2.44 | #97 |
| 6 | Step 3.5 Flash | 0 | #270 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=matharena-arxivlean-march · How It Works · Data refreshed daily, snapshot 2026-10-20.