Open Arabic LLM - Arabic MMLU Computer Science (University): leaderboard
Metric: Accuracy (%). Source: huggingface.co. 163 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen2.5-32B-Instruct-CFT | 82.81 |
| 2 | lambda-qwen2.5-14B-dpo-test | 81.25 |
| 3 | Awqward2.5-32B-Instruct | 81.25 |
| 4 | Qwen 2.5 32B Instruct | 79.69 |
| 5 | Qwen 2.5 72B Instruct | 78.12 |
| 6 | Qwen2.5-32B-Instruct-abliterated-v2 | 78.12 |
| 7 | Qwentile2.5-32B-Instruct | 78.12 |
| 8 | Saka-14B | 78.12 |
| 9 | Qwen2.5-Lumen-14B | 78.12 |
| 10 | Qwen3-8B-Base | 76.56 |
| 11 | EZO-Qwen2.5-32B-Instruct | 76.56 |
| 12 | free-evo-qwen72B-v0.8-re | 76.56 |
| 13 | Llama 3.3 70B Instruct | 75 |
| 14 | Qwen 2.5 14B | 75 |
| 15 | gemma-3-12B-pt | 73.44 |
Interactive version: theaggregate.ai/benchmark?slug=open-arabic-llm-arabic-mmlu-computer-science-university · How It Works · Data refreshed daily, snapshot 2026-09-05.