ArXivMath 02/2026
Math · 2026-03-13
This MathArena track evaluates models on the ArXivMath 02/2026 problem set. Scores report the percentage of problems answered correctly.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.4 Pro | 75.8 |
| GPT-5.5 | 73.4 |
| Gemini 3.1 Pro Preview | 62.5 |
| Opus 4.8 | 60.2 |
| GPT-5.4 | 59.4 |
| DeepSeek-V4-Pro | 51.6 |
| Gemini 3.5 Flash | 50.8 |
| GLM-5.2 | 46.1 |
| DeepSeek-V4-Flash | 43.8 |
| Kimi K2.6 | 43.0 |
| GLM-5 | 41.4 |
| Opus 4.6 | 40.6 |
| Opus 4.7 | 40.6 |
| GLM-5.1 | 39.1 |
| GPT-5.2 | 37.5 |
| Step 3.5 Flash | 32.8 |
| Grok 4.1 Fast | 32.0 |
| Step 3.7 Flash | 32.0 |
| Qwen3.5-27B | 31.3 |
| Nemotron 3 Super 120B A12B | 30.5 |
| Qwen3.5 35B A3B | 30.5 |
| Qwen3.5-9B | 27.3 |
| Qwen3 30B A3B Thinking 2507 | 22.7 |
| Qwen3.5-4B | 19.4 |
| Qwen3 4B Thinking 2507 | 18.0 |
| QED-Nano | 14.1 |