MMLU-by-task - High School Computer Science: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 1257 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 2 70B Base | 79 |
| 2 | Llama 2 70B Chat | 77 |
| 3 | Mistral-7B-v0.1 | 68 |
| 4 | LLaMA-65B | 68 |
| 5 | Llama 2 70B Chat (HF) | 65 |
| 6 | vicuna-33B-v1.3 | 63 |
| 7 | LLaMA-30B | 60 |
| 8 | Llama 2 13B Chat Base | 59 |
| 9 | vicuna-13B-v1.5 | 56 |
| 10 | MythoMax-L2-13B | 56 |
| 11 | vicuna-13B-v1.3 | 55 |
| 12 | mpt-30B | 53 |
| 13 | mpt-30B Instruct | 51 |
| 14 | LLaMA-13B | 49 |
| 15 | CodeLlama-13B-Instruct-hf | 49 |
Interactive version: theaggregate.ai/benchmark?slug=mmlu-by-task-high-school-computer-science · How It Works · Data refreshed daily, snapshot 2026-09-05.