HuggingFace Open LLM Leaderboard: leaderboard
The definitive open-source model leaderboard. Tests on IFEval, BBH, MATH, GPQA, MUSR, and MMLU-PRO. Community-driven and continuously updated.
Metric: Average score. Source: huggingface.co. Status: saturated. 4576 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | calme-3.2-instruct-78b | 52.08 |
| 2 | calme-3.1-instruct-78b | 51.29 |
| 3 | CalmeRys-78B-Orpo-v0.1 | 51.23 |
| 4 | calme-2.4-rys-78b | 50.77 |
| 5 | Qwen2.5-72B-Instruct-abliterated | 48.11 |
| 6 | Qwen2.5-72B-Instruct | 47.98 |
| 7 | calme-2.1-qwen2.5-72b | 47.86 |
| 8 | Homer-v1.0-Qwen2.5-72B | 47.46 |
| 9 | qwen2.5-test-32b-it | 47.37 |
| 10 | Linkbricks-Horizon-AI-Avengers-V1-32B | 47.34 |
Interactive version: theaggregate.ai/benchmark?slug=huggingface-open-llm-leaderboard · How It Works · Data refreshed daily, snapshot 2026-09-05.