MMLU Pro - Health
General QA · 2024-06-03
The Health 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 Opus 5 | 87.8 |
| Gemini 3.1 Pro Preview | 86.4 |
| Opus 4.8 | 85.1 |
| Opus 4.1 | 85.0 |
| Claude Fable 5 | 84.6 |
| Opus 4.5 | 84.6 |
| Opus 4.6 | 84.6 |
| Gemini 3.6 Flash | 84.6 |
| GPT-5.5 | 84.4 |
| Qwen3.7-Max | 84.4 |
| GPT-5.6 Sol | 84.1 |
| Opus 4 | 84.0 |
| Grok 4.5 | 84.0 |
| MiniMax M2.1 | 83.7 |
| Gemini 3.5 Flash | 83.6 |
| Gemini 3 Pro Preview | 83.6 |
| Muse Spark 1.1 | 83.4 |
| GPT-5.6 Terra | 82.8 |
| Gemini 3 Flash Preview | 82.6 |
| Grok 4.3 | 82.6 |
| Opus 4.7 | 82.4 |
| Sonnet 4.6 | 82.4 |
| DeepSeek-V4-Pro | 82.4 |
| GPT-5.4 | 82.4 |
| Sonnet 5 | 82.3 |
| Sonnet 4.5 | 82.0 |
| Kimi K3 | 82.0 |
| Grok 4 | 81.7 |
| GLM-5.1 | 81.5 |
| GLM-5.2 | 81.4 |
| GPT-5.6 Luna | 81.4 |
| GPT-5.2 | 81.2 |
| GPT-5.1 | 80.9 |
| Grok 4.20 | 80.9 |
| GPT-5 | 80.8 |
| Kimi K2.5 | 80.8 |
| Muse Spark | 80.8 |
| Qwen3.6 Plus (2026-04-02) | 80.8 |
| Sonnet 4 | 80.7 |
| Kimi K2.6 | 80.7 |