MMLU-by-task Leaderboard: leaderboard
MMLU-by-task leaderboard exposing per-subject MMLU performance plus an average score across many open-weight models.
Metric: MMLU Average (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 1257 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 2 70B Base | 69.83 | #1025 |
| 2 | Xwin-LM-70B-V0.1 | 69.77 | #982 |
| 3 | Llama 2 70B Chat | 69.18 | #1237 |
| 4 | StableBeluga2 | 68.79 | #1078 |
| 5 | Mistral-7B-v0.1 | 64.16 | #1326 |
| 6 | WizardLM-70B-V1.0 | 64.05 | #1236 |
| 7 | Llama 2 70B Chat (HF) | 63.91 | #1264 |
| 8 | internlm-20B | 61.85 | #1296 |
| 9 | vicuna-33B-v1.3 | 59.22 | #1347 |
| 10 | vicuna-13B-v1.5 | 56.67 | #1392 |
| 11 | OpenHermes-13B | 56.35 | #1393 |
| 12 | falcon-40B Instruct | 55.45 | #1458 |
| 13 | MythoMax-L2-13B | 55.33 | #1409 |
| 14 | trurl-2-13B-academic | 55.2 | #1413 |
| 15 | Chinese-Alpaca-2-13B | 55.1 | #1414 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmlu-by-task-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.