MMLU-by-task - High School Mathematics: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 1257 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 2 70B Base | 35.56 |
| 2 | LLaMA-65B | 34.07 |
| 3 | Mistral-7B-v0.1 | 33.7 |
| 4 | mpt-30B Instruct | 32.96 |
| 5 | MythoMax-L2-13B | 32.59 |
| 6 | vicuna-13B-v1.5 | 31.85 |
| 7 | Llama 2 13B Chat Base | 31.11 |
| 8 | CodeLlama-34B-Instruct-hf | 31.11 |
| 9 | Llama 2 7B Base | 30.37 |
| 10 | Llama 2 70B Chat (HF) | 30 |
| 11 | CodeLlama-7B-hf | 29.63 |
| 12 | stablelm-base-alpha-7B | 29.26 |
| 13 | mpt-30B | 28.52 |
| 14 | Llama 2 70B Chat | 27.78 |
| 15 | gpt-j-6B | 27.41 |
Interactive version: theaggregate.ai/benchmark?slug=mmlu-by-task-high-school-mathematics · How It Works · Data refreshed daily, snapshot 2026-09-05.