MMLU-by-task - Conceptual Physics — leaderboard
Metric: Accuracy (%). Source: huggingface.co. 1257 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 2 70B Base | 66.38 |
| 2 | falcon-180B | 66.38 |
| 3 | StableBeluga2 | 65.11 |
| 4 | Llama 2 70B Chat | 63.4 |
| 5 | Llama 2 70B Chat (HF) | 58.3 |
| 6 | LLaMA-65B | 58.3 |
| 7 | Mistral-7B-v0.1 | 57.45 |
| 8 | Llama 2 70B Chat GPTQ | 53.62 |
| 9 | LLaMA-30B | 51.49 |
| 10 | MythoMax-L2-13B | 47.66 |
| 11 | OpenHermes-13B | 47.66 |
| 12 | vicuna-33B-v1.3 | 47.23 |
| 13 | mpt-30B Instruct | 46.38 |
| 14 | vicuna-7B-v1.5 | 45.11 |
| 15 | WizardCoder-Python-34B-V1.0 | 45.11 |
Interactive version: theaggregate.ai/benchmark?slug=mmlu-by-task-conceptual-physics · How the rankings work · Data refreshed daily, snapshot 2026-07-22.