BRUMO 2025
Math · 2025-05-29
This MathArena track evaluates models on the BRUMO 2025 problem set. Scores report the percentage of problems answered correctly.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3 Flash Preview | 100.0 |
| Step 3.5 Flash | 100.0 |
| DeepSeek-V3.2-Speciale | 99.2 |
| GLM-5 | 99.2 |
| Gemini 3 Pro Preview | 98.3 |
| GPT-5.2 | 98.3 |
| Kimi K2.5 | 98.3 |
| Grok 4.1 Fast | 97.5 |
| DeepSeek-V3.2 | 96.7 |
| DeepSeek-V3.2-Exp | 95.8 |
| Grok 4 Fast | 95.8 |
| o3 | 95.8 |
| Grok 4 | 95.0 |
| GLM 4.6 | 94.2 |
| GPT-5.1 | 93.3 |
| Kimi K2 Thinking | 93.3 |
| DeepSeek-R1-0528 | 92.5 |
| GLM-4.5 | 92.5 |
| gpt-oss-120b | 92.5 |
| GPT-5 | 91.7 |
| Sonnet 4.5 | 90.8 |
| DeepSeek-V3.1 | 90.0 |
| Gemini 2.5 Pro | 90.0 |
| GLM 4.5 Air | 90.0 |
| GPT-5 Mini | 90.0 |
| Gemini 2.5 Pro Preview 05-06 | 89.2 |
| QED-Nano | 87.5 |
| gpt-oss-20b | 86.7 |
| O4 Mini | 86.7 |
| Qwen3-235B-A22B | 86.7 |
| Falcon H1R 7B | 85.8 |
| Grok 3 Mini | 85.0 |
| Gemini 2.5 Flash | 83.3 |
| K2-Think | 83.3 |
| Opus 4 | 81.7 |
| DeepSeek-R1 | 80.8 |
| GPT-5 Nano | 80.8 |
| Qwen3-30B-A3B | 77.5 |
| DeepSeek R1 Distill Qwen 14B | 68.3 |
| DeepSeek-R1-Distill-Qwen-32B | 68.3 |