ArXivMath 05/2026
Math · 2026-06-11
This MathArena track evaluates models on the ArXivMath 05/2026 problem set. Published scores average correctness across observed problem attempts; the item count records distinct problems, not a single-run binomial denominator.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6.1 Sol | 95.0 |
| GPT-6 Astra | 95.0 |
| GPT-6 Sol | 90.0 |
| Claude Fable 5.1 | 87.5 |
| Claude Fable 5 | 87.5 |
| Claude Opus 5.5 | 81.7 |
| GPT-5.5 | 77.5 |
| Muse Spark 1.3 | 73.3 |
| Opus 4.8 | 65.0 |
| Kimi K3 | 61.7 |
| Grok 4.7 | 60.8 |
| DeepSeek V4.1 Flash | 58.3 |
| Gemini 3.8 Flash | 58.3 |
| GLM-5.2 | 56.7 |
| DeepSeek-V4-Flash | 55.8 |
| Gemini 3.6 Flash | 54.2 |
| GLM-5.1 | 52.5 |
| Gemini 3.1 Pro Preview | 51.7 |
| Gemini 3.7 Flash | 47.5 |
| Step 3.7 Flash | 46.7 |
| Gemini 3.5 Flash | 44.2 |
| Qwen3.6 35B A3B | 37.5 |
| Qwen3.5-2B | 10.6 |