MMLU Pro - Math
Math · 2024-06-03
The Math split of MMLU Pro. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 97.6 |
| Claude Opus 5 | 97.1 |
| Gemini 3.5 Flash | 96.3 |
| Gemini 3.1 Pro Preview | 96.0 |
| Gemini 3.6 Flash | 96.0 |
| Gemini 3 Pro Preview | 96.0 |
| Muse Spark 1.1 | 96.0 |
| Qwen3.7-Max | 96.0 |
| Opus 4.6 | 95.9 |
| Sonnet 5 | 95.9 |
| Grok 4.5 | 95.9 |
| Opus 4.7 | 95.9 |
| Opus 4.8 | 95.9 |
| GLM-5.2 | 95.9 |
| Gemini 3 Flash Preview | 95.7 |
| Kimi K2.5 | 95.7 |
| Kimi K2.6 | 95.6 |
| Kimi K3 | 95.6 |
| Nemotron 3 Ultra 550B A55B | 95.6 |
| Qwen3.5 Plus (2026-02-15) | 95.6 |
| Qwen3 Max Thinking (2026-01-23) | 95.6 |
| GLM-5.1 | 95.5 |
| Qwen3.6 Plus (2026-04-02) | 95.5 |
| DeepSeek-V4-Pro | 95.3 |
| GPT-5.6 Sol | 95.3 |
| Grok 4.20 | 95.1 |
| DeepSeek-V3.2 | 95.0 |
| GPT-5.2 | 95.0 |
| Muse Spark | 95.0 |
| Qwen3 Max (rolling alias) | 95.0 |
| Sonnet 4.5 | 95.0 |
| Inkling | 94.9 |
| Qwen3.5-Flash | 94.9 |
| Gemini 3.1 Flash-Lite Preview | 94.8 |
| GLM-5 | 94.8 |
| GPT-5.5 | 94.7 |
| GPT-5 | 94.7 |
| GPT-5.4 | 94.7 |
| Opus 4.1 | 94.6 |
| MiniMax M3 | 94.6 |