MMLU-by-task - TruthfulQA MC1: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 1253 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 2 70B Chat | 40.27 |
| 2 | vicuna-33B-v1.3 | 36.96 |
| 3 | MythoMax-L2-13B | 36.6 |
| 4 | vicuna-13B-v1.3 | 36.47 |
| 5 | vicuna-13B | 36.11 |
| 6 | Llama 2 70B Chat (HF) | 35.62 |
| 7 | vicuna-13B-v1.5 | 35.62 |
| 8 | vicuna-7B-v1.3 | 31.7 |
| 9 | Llama 2 70B Base | 31.09 |
| 10 | Llama 2 7B Chat | 30.11 |
| 11 | trurl-2-7B | 30.11 |
| 12 | ChatGLM2 6B | 29.99 |
| 13 | CodeLlama-13B-Instruct-hf | 29.38 |
| 14 | CodeLlama-34B-Instruct-hf | 29.01 |
| 15 | falcon-7B Instruct | 28.89 |
Interactive version: theaggregate.ai/benchmark?slug=mmlu-by-task-truthfulqa-mc1 · How It Works · Data refreshed daily, snapshot 2026-09-05.