Open Arabic LLM - Arabic MMLU HT Logical Fallacies — leaderboard
Metric: Accuracy (%). Source: huggingface.co. 163 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 76.69 |
| 2 | Ultiima-72B | 76.69 |
| 3 | Rombos-LLM-V2.5-Qwen-72B | 76.07 |
| 4 | calme-2.1-qwen2.5-72B | 74.23 |
| 5 | calme-2.2-qwen2.5-72B | 73.62 |
| 6 | Qwen 2 72B | 73.01 |
| 7 | Gemma 3 27B (IT) | 71.17 |
| 8 | Qwentile2.5-32B-Instruct | 71.17 |
| 9 | gemma-3-27B-pt | 70.55 |
| 10 | Awqward2.5-32B-Instruct | 69.33 |
| 11 | lambda-qwen2.5-32B-dpo-test | 69.33 |
| 12 | Llama 3.3 70B Instruct | 68.71 |
| 13 | Qwen 2.5 72B | 68.71 |
| 14 | Qwen2.5-32B-Instruct-abliterated-v2 | 68.71 |
| 15 | Gemma 2 27B (IT) | 68.1 |
Interactive version: theaggregate.ai/benchmark?slug=open-arabic-llm-arabic-mmlu-ht-logical-fallacies · How the rankings work · Data refreshed daily, snapshot 2026-07-22.