Benchmarks / Math
ArXivMath 06/2026 (MathArena)
Research-level math problems with a checkable final answer, drawn from arXiv papers submitted in June 2026; each edition is restricted to papers published within the last month to reduce training-data contamination. Maintained by MathArena. Score: Accuracy (average performance on the competition) with a 95% confidence interval (normal approximation).
ArXivMath 06/2026 (MathArena)
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GPT-6.1 Sol | max | 94.4% |
| 2 | GPT-6 Astra | max | 94.4% |
| 3 | GPT-6 Sol | max | 91.0% |
| 4 | Fable 5.1 | max | 91.0% |
| 5 | GPT-5.6 Sol | max | 88.5% |
| 6 | Fable 5 | max | 85.4% |
| 7 | GPT-5.5 | xhigh | 83.6% |
| 8 | Opus 5.5 | high | 83.3% |
| 9 | Opus 5 | max | 82.6% |
| 10 | Muse Spark 1.3 | max | 76.7% |
| 11 | Qwen 3.8 Max | max | 75.7% |
| 12 | Kimi K3 | max | 71.5% |
| 13 | Grok 4.7 | xhigh | 70.1% |
| 14 | Opus 4.8 | max | 70.0% |
| 15 | Gemini 3.8 Flash | max | 67.4% |
| 16 | Gemini 3.1 Pro | high | 66.7% |
| 17 | Grok 4.5 | max | 66.7% |
| 18 | Gemini 3.7 Flash | max | 62.5% |
| 19 | DeepSeek V4.1 Flash | max | 57.6% |
| 20 | Gemini 3.6 Flash | xhigh | 56.9% |
| 21 | Muse Spark 1.1 | max | 56.9% |
| 22 | Gemini 3.5 Flash | xhigh | 52.1% |
| 23 | GLM 5.2 | max | 46.5% |
| 24 | DeepSeek V4 Flash | max | 43.8% |
| 25 | Qwen 3.6 35B | – | 27.1% |
| 26 | Qwen 3.5 2B | low | 4.2% |