ArXivMath 01/2026
Math · 2026-02-04
This MathArena track evaluates models on the ArXivMath 01/2026 problem set. Scores report the percentage of problems answered correctly.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.4 | 76.1 |
| DeepSeek-V4-Pro | 73.9 |
| GPT-5.5 | 73.9 |
| Opus 4.6 | 72.8 |
| Kimi K2.6 | 71.7 |
| Gemini 3.1 Pro Preview | 70.7 |
| GPT-5.2 | 67.9 |
| GLM-5.1 | 65.2 |
| Kimi K2.5 | 62.5 |
| Gemini 3 Pro Preview | 62.0 |
| Step 3.5 Flash | 60.3 |
| Gemini 3 Flash Preview | 58.1 |
| DeepSeek-V3.2 | 57.1 |
| Qwen3.5 397B A17B | 54.9 |
| GLM-5 | 54.4 |
| Grok 4.1 Fast | 53.3 |
| Qwen3.5-27B | 53.3 |
| Opus 4.7 | 52.2 |
| Qwen3.5 35B A3B | 50.0 |
| Nemotron 3 Super 120B A12B | 48.9 |
| Qwen3.5-9B | 44.6 |
| Qwen3 30B A3B Thinking 2507 | 39.1 |
| Qwen3.5-4B | 38.1 |
| QED-Nano | 37.0 |
| Qwen3 4B Thinking 2507 | 23.9 |