MathArena - CMIMC 2025: leaderboard
Carnegie Mellon Informatics and Mathematics Competition 2025: mathematics and computer science problems including combinatorics, algebra, geometry, and algorithms.
Metric: Accuracy (%). Source: matharena.ai. Saturation forecast: Estimated already saturated. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (High) | 95.62 | #110 (GPT-5.2) |
| 2 | Step 3.5 Flash | 95.62 | #270 |
| 3 | DeepSeek V3.2 Speciale | 95.62 | #128 |
| 4 | GPT-5.1 (High) | 94.38 | #135 (GPT-5.1) |
| 5 | GLM-5 | 93.12 | #141 |
| 6 | Kimi K2 (Thinking) | 90.62 | #243 (Kimi K2) |
| 7 | Kimi K2.5 (Thinking) | 90.62 | #144 (Kimi K2.5) |
| 8 | Gemini 3 Flash | 90 | #99 |
| 9 | Gemini 3 Pro (Preview) | 90 | #66 |
| 10 | GPT-5 (High) | 88.75 | #98 (GPT-5) |
| 11 | GLM-4.6 | 87.5 | #252 |
| 12 | Grok 4 | 85 | #181 |
| 13 | GPT-OSS-120B (High) | 85 | #341 (GPT-OSS-120B) |
| 14 | Grok 4 Fast (Reasoning) | 85 | #254 (Grok 4 Fast) |
| 15 | Grok 4.1 Fast (Reasoning) | 84.38 | #217 (Grok 4.1 Fast) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=matharena-cmimc-2025 · How It Works · Data refreshed daily, snapshot 2026-10-20.