Open LLM Leaderboard v1 - MMLU: leaderboard
Original Hugging Face Open LLM Leaderboard academic knowledge across the 57 MMLU subjects, with five demonstrations and the published subject-aggregate accuracy. Percentage, higher is better. Frozen v1 archive, distinct from MMLU-Pro in v2. Public test questions may overlap model training.
Metric: Accuracy (%) (5-shot). Source: huggingface.co. Saturation forecast: Estimated already saturated. 7252 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | 34B-beta | 85.6 | #895 |
| 2 | Qwen 2 72B | 83.79 | #517 |
| 3 | Llama 3 70B Japanese Suzume Vector V0.1 | 80.28 | #774 |
| 4 | Qwen 1.5 110B | 80.2 | #735 |
| 5 | autotrain-llama3-70B-orpo-v1 | 80.11 | #844 |
| 6 | Llama 3 70B Instruct | 80.06 | #643 |
| 7 | Llama3-TenyxChat-70B | 80.04 | #739 |
| 8 | Llama 3 70B | 79.23 | #635 |
| 9 | Llama 3 70B Fireplace | 78.97 | #847 |
| 10 | YiSM-34B-0rn | 78.51 | #715 |
| 11 | Yi-1.5-34B-32K | 78.15 | #1107 |
| 12 | blossom-v5.1-34B | 78.05 | #782 |
| 13 | Qwen 1.5 110B Chat | 78.04 | #695 |
| 14 | Yi 1.5 34B | 78 | #808 |
| 15 | free-evo-qwen72B-v0.8-re | 78 | #442 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-llm-leaderboard-v1-mmlu · How It Works · Data refreshed daily, snapshot 2026-11-07.