Massive Multitask Language Understanding — 57 subjects across STEM, humanities, and social sciences at varying difficulty levels from elementary to professional.
The most widely reported knowledge benchmark. Broad coverage across many domains. Extensive historical baselines.
Heavily saturated (top models >90%). Significant contamination risk. MMLU-Pro is the harder successor.
Higher is better. Accuracy percentage (0–100%). Random baseline ~25% (4-choice).
Heavily saturated (top models >90%).
| # | Model | Score |
|---|---|---|
| 1 | 94.1 | |
| 2 | 94.0 | |
| 3 | 94.0 | |
| 4 | 94.0 | |
| 5 | 93.9 | |
| 6 | 93.4 | |
| 7 | 93.3 | |
| 8 | 93.1 | |
| 9 | 93.0 | |
| 10 | 92.6 | |
| 11 | 92.6 | |
| 12 | 92.5 | |
| 13 | 92.0 | |
| 14 | 92.0 | |
| 15 | 92.0 | |
| 16 | 91.9 | |
| 17 | 91.8 | |
| 18 | 91.8 | |
| 19 | 91.8 | |
| 20 | 91.8 | |
| 21 | 91.7 | |
| 22 | 91.6 | |
| 23 | 91.6 | |
| 24 | 91.6 | |
| 25 | 91.3 | |
| 26 | 91.3 | |
| 27 | 91.2 | |
| 28 | 91.2 | |
| 29 | 91.1 | |
| 30 | 91.1 | |
| 31 | 91.0 | |
| 32 | 91.0 | |
| 33 | 91.0 | |
| 34 | 91.0 | |
| 35 | 91.0 | |
| 36 | 91.0 | |
| 37 | 91.0 | |
| 38 | 90.9 | |
| 39 | 90.9 | |
| 40 | 90.9 | |
| 41 | 90.8 | |
| 42 | 90.8 | |
| 43 | 90.8 | |
| 44 | 90.8 | |
| 45 | 90.8 | |
| 46 | 90.8 | |
| 47 | 90.8 | |
| 48 | 90.8 | |
| 49 | 90.8 | |
| 50 | 90.7 |