Omni-MATH: leaderboard
Olympiad-level mathematics benchmark spanning domains and difficulty bands, using structured grading to evaluate advanced mathematical reasoning.
Metric: Overall Accuracy (%). Source: omni-math.github.io. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 Mini | 60.54 | #323 |
| 2 | O1 Preview | 52.55 | #193 |
| 3 | Qwen2.5-Math-72B-Instruct | 36.2 | #717 |
| 4 | Qwen2.5-Math-7B-Instruct | 33.22 | #1444 |
| 5 | GPT-4o | 30.49 | #343 |
| 6 | Qwen2-Math-7B-Instruct | 29.36 | #1370 |
| 7 | Claude 3.5 Sonnet | 26.23 | #350 |
| 8 | Llama 3.1 70B Instruct | 24.16 | #561 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=omni-math · How It Works · Data refreshed daily, snapshot 2026-10-20.