HMMT 2025: leaderboard
MathArena evaluation based on Harvard-MIT Mathematics Tournament 2025 problems, emphasizing olympiad-style high-school contest reasoning.
Metric: Score (%). Source: matharena.ai. Source: llm-stats.com. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 Pro | 100 | #61 |
| 2 | GPT-5.2 | 99.4 | #110 |
| 3 | DeepSeek V3.2 Speciale | 99.2 | #128 |
| 4 | Qwen 3.6 Plus | 96.7 | #120 |
| 5 | Kimi K2.5 | 95.4 | #144 |
| 6 | Qwen 3.5 397B A17B | 94.8 | #146 |
| 7 | GLM-5.2 | 94.4 | #73 |
| 8 | GLM-5.1 | 94 | #97 |
| 9 | Qwen 3.6 27B | 93.8 | #170 |
| 10 | GPT-5 | 93.3 | #98 |
| 11 | Grok 4 Fast | 93.3 | #254 |
| 12 | Qwen 3.5 27B | 92 | #187 |
| 13 | Qwen 3.5 122B A10B | 91.4 | #173 |
| 14 | Qwen 3.6 35B A3B | 90.7 | #226 |
| 15 | DeepSeek V3.2 | 90.2 | #205 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=hmmt-2025 · How It Works · Data refreshed daily, snapshot 2026-10-20.