MathArena - HMMT Feb 2026: leaderboard
Harvard-MIT Mathematics Tournament February 2026, evaluated close to competition date for contamination-free olympiad-level math assessment.
Metric: Accuracy (%). Source: matharena.ai. Saturation forecast: Estimated already saturated. 31 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 98.48 | #31 (GPT-5.5) |
| 2 | GPT-5.4 (xHigh) | 97.73 | #79 (GPT-5.4) |
| 3 | GPT-5.2 (High) | 96.97 | #110 (GPT-5.2) |
| 4 | Kimi K3 (Thinking) | 96.97 | #30 (Kimi K3) |
| 5 | Claude Opus 4.6 (High) | 96.21 | #63 (Claude Opus 4.6) |
| 6 | Gemini 3.5 Flash | 95.45 | #60 |
| 7 | Claude Opus 4.8 (Max) | 95.45 | #37 (Claude Opus 4.8) |
| 8 | Gemini 3.1 Pro (Preview) | 94.7 | #58 |
| 9 | Kimi K2.6 (Thinking) | 94.7 | #104 (Kimi K2.6) |
| 10 | DeepSeek V4 Pro (Max) | 93.94 | #102 (DeepSeek V4 Pro) |
| 11 | DeepSeek V4 Flash (Max) | 93.94 | #137 (DeepSeek V4 Flash) |
| 12 | Claude Opus 4.7 (xHigh) | 93.94 | #49 (Claude Opus 4.7) |
| 13 | GLM-5.2 | 92.42 | #73 |
| 14 | Gemini 3 Flash | 89.39 | #99 |
| 15 | GLM-5.1 | 89.39 | #97 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=matharena-hmmt-feb-2026 · How It Works · Data refreshed daily, snapshot 2026-10-20.