Benchmarks · math
MathArena
Uses MathArena's website-wide IRT expected-performance estimate. Score: website-wide IRT expected performance. Collected by Jake Boggs' benchmark dashboard.
84 models · 91 results · data as of 2026-10-29 · matharena.ai ↗
Leaderboard
GPT-6.1 Sol max
All results
91 rows| # | Model | Effort | MathArena |
|---|---|---|---|
| 1 | GPT-6.1 Sol | max | 84.0% |
| 2 | GPT-6 Astra | max | 81.9% |
| 3 | GPT-6 Sol | max | 80.1% |
| 4 | Opus 5.5 | max | 79.0% |
| 5 | Fable 5.1 | max | 78.1% |
| 6 | GPT-6 Astra | low | 77.3% |
| 7 | Opus 5 | max | 70.4% |
| 8 | Gemini 4 Argon | max | 69.3% |
| 9 | GPT-5.6 Sol | max | 62.9% |
| 10 | GPT-5.5 | xhigh | 61.3% |
| 11 | GPT-5.4 Pro | xhigh | 52.5% |
| 12 | Qwen 3.8 Max | max | 51.6% |
| 13 | Fable 5 | max | 51.5% |
| 14 | Kimi K3 | max | 48.0% |
| 15 | Muse Spark 1.3 | max | 43.0% |
| 16 | Muse Spark 1.1 | max | 42.5% |
| 17 | DeepSeek V4.1 Flash | max | 42.3% |
| 18 | Grok 4.7 | xhigh | 41.1% |
| 19 | Opus 4.8 | max | 39.8% |
| 20 | GPT-5.4 | xhigh | 37.2% |
| 21 | Grok 4.5 | max | 34.1% |
| 22 | Gemini 3.8 Flash | max | 34.0% |
| 23 | GPT-5.2 | xhigh | 33.7% |
| 24 | Qwen 3.8 | xhigh | 31.6% |
| 25 | Gemini 3.1 Pro | high | 30.1% |
| 26 | Mistral Large 4 | high | 26.7% |
| 27 | DeepSeek V4 Pro | max | 26.2% |
| 28 | Gemini 3.7 Flash | max | 25.9% |
| 29 | DeepSeek V4 Flash | max | 24.8% |
| 30 | Gemini 3.5 Flash | xhigh | 24.7% |