Vellum - AIME 2025: leaderboard
Vellum's independently-run AIME 2025 evaluation. American Invitational Mathematics Examination problems.
Metric: Accuracy (%). Source: www.vellum.ai. Saturation forecast: Estimated already saturated. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 100 | #110 |
| 2 | Gemini 3 Pro | 100 | #80 |
| 3 | Best Expert | 100 | |
| 4 | Claude Opus 4.6 | 99.8 | #63 |
| 5 | Kimi K2 (Thinking) | 99.1 | #243 (Kimi K2) |
| 6 | GPT-OSS-20B | 98.7 | #518 |
| 7 | O3 | 98.4 | #124 |
| 8 | GPT-OSS-120B | 97.9 | #341 |
| 9 | Claude Haiku 4.5 | 96.3 | #280 |
| 10 | Kimi K2.5 | 96.1 | #144 |
| 11 | GPT-5.1 | 94 | #135 |
| 12 | Grok 3 | 93.3 | #304 |
| 13 | O4 Mini | 92.7 | #180 |
| 14 | Grok 4 | 91.7 | #181 |
| 15 | GPT-5 | 91 | #98 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vellum-aime-2025 · How It Works · Data refreshed daily, snapshot 2026-10-20.