Open Arabic LLM - Arabic MMLU HT Security Studies: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 163 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemma 3 27B (IT) | 76.73 |
| 2 | Saka-14B | 75.1 |
| 3 | tempmotacilla-cinerea-0308 | 73.06 |
| 4 | free-evo-qwen72B-v0.8-re | 72.24 |
| 5 | Qwentile2.5-32B-Instruct | 71.84 |
| 6 | Qwen2.5-32B-Instruct-abliterated-v2 | 71.43 |
| 7 | Qwen 2.5 32B Instruct | 70.61 |
| 8 | Awqward2.5-32B-Instruct | 70.61 |
| 9 | Qwen2.5-32B-Instruct-CFT | 70.2 |
| 10 | Qwen 2.5 72B Instruct | 69.8 |
| 11 | gemma-3-27B-pt | 69.39 |
| 12 | Llama 3.3 70B Instruct | 68.98 |
| 13 | Qwen3-8B-Base | 68.98 |
| 14 | Qwen2.5-Lumen-14B | 68.57 |
| 15 | Qwen2.5-14B-Gutenberg-1e-Delta | 68.57 |
Interactive version: theaggregate.ai/benchmark?slug=open-arabic-llm-arabic-mmlu-ht-security-studies · How It Works · Data refreshed daily, snapshot 2026-09-05.