Open Persian LLM - MMLU Pro — leaderboard

Metric: Accuracy (%). Source: huggingface.co. 31 models tracked.

Top models

#ModelScore
1Llama 3.3 70B Instruct43.67
2Gemma 3 27B (IT)40.1
3Qwen 2.5 32B Instruct38.19
4QwQ 32B-Preview37.94
5Gemma 2 27B (IT)36.28
6Gemma 2 9B (IT)31.74
7aya-expanse-32B31.03
8Gemma 3 12B (IT)29.36
9Llama 3 8B Instruct25.54
10aya-23-35B25.54
11aya-expanse-8B25.06
12Llama 3.1 8B24.1
13Qwen 2 7B Instruct23.87
14Qwen 2.5 7B Instruct21.96
15suzume-llama-3-8B-multilingual21.95

Interactive version: theaggregate.ai/benchmark?slug=open-persian-llm-mmlu-pro · How the rankings work · Data refreshed daily, snapshot 2026-07-22.