Open Arabic LLM - Arabic MMLU Average: leaderboard
The "Arabic MMLU Average" subtask of Open Arabic LLM. A task from the Open Arabic LLM Leaderboard, evaluating models on Arabic-language knowledge, reasoning and understanding across native and translated benchmarks.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 71.43 | #454 |
| 2 | Saka-14B | 70.84 | #344 |
| 3 | gemma-3-27B-pt | 70.76 | #502 |
| 4 | AIC-1 | 69.94 | #231 |
| 5 | Llama 3.3 70B Instruct | 69.85 | #534 |
| 6 | Qwen2.5-32B-Instruct-abliterated-v2 | 69.58 | #394 |
| 7 | AceGPT-v2-32B-Chat | 68.92 | #507 |
| 8 | Qwen 2.5 14B | 68.49 | #725 |
| 9 | Qwentile2.5-32B-Instruct | 68.24 | #339 |
| 10 | Awqward2.5-32B-Instruct | 68.08 | #371 |
| 11 | tempmotacilla-cinerea-0308 | 68.07 | #395 |
| 12 | Qwen 2.5 32B Instruct | 68.03 | #498 |
| 13 | Qwen2.5-32B-Instruct-CFT | 67.96 | #375 |
| 14 | free-evo-qwen72B-v0.8-re | 67.77 | #435 |
| 15 | lambda-qwen2.5-14B-dpo-test | 67.22 | #481 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-arabic-llm-arabic-mmlu-average · How It Works · Data refreshed daily, snapshot 2026-10-20.