Benchmarks · math
MATH-500
MATH-500: competition math subset; accuracy.
34 models · 38 results · data as of 2026-10-29 · artificialanalysis.ai ↗
Leaderboard
GPT-5 high
All results
38 rows| # | Model | Effort | MATH-500 |
|---|---|---|---|
| 1 | GPT-5 | high | 99.4% |
| 2 | o3 | 99.2% | |
| 3 | GPT-5 | medium | 99.1% |
| 4 | Grok 4 | 99.0% | |
| 5 | o4 mini | high | 98.9% |
| 6 | GPT-5 | low | 98.7% |
| 7 | o3 mini | high | 98.5% |
| 8 | GLM 4.5 | 97.9% | |
| 9 | o3 mini | 97.3% | |
| 10 | Kimi K2 | none | 97.1% |
| 11 | o1 | 97.0% | |
| 12 | Gemini 2.5 Pro | 96.7% | |
| 13 | GLM 4.5 Air | 96.5% | |
| 14 | o1 mini | 94.4% | |
| 15 | Gemini 2.5 Flash | none | 93.2% |
| 16 | Gemini 2.5 Flash Lite | none | 92.6% |
| 17 | GPT-4.1 Mini | none | 92.5% |
| 18 | GPT-4.1 | none | 91.3% |
| 19 | Mistral Medium 3 | none | 90.7% |
| 20 | Llama 4 Maverick | none | 88.9% |
| 21 | Mistral Small 3.2 | none | 88.3% |
| 22 | Grok 3 | none | 87.0% |
| 23 | GPT-5 | minimal | 86.1% |
| 24 | GPT-4.1 Nano | none | 84.8% |
| 25 | Llama 4 Scout | none | 84.4% |
| 26 | Qwen 2.5 Max | none | 83.5% |
| 27 | Command A | none | 81.9% |
| 28 | GPT-4o mini | none | 78.9% |
| 29 | Mistral Small 3 | none | 71.5% |
| 30 | Devstral Medium | none | 70.7% |