Open Arabic LLM - Arabic MMLU HT Management: leaderboard
The "Arabic MMLU HT Management" 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: Estimated already saturated. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 77.67 | #454 |
| 2 | AceGPT-v2-32B-Chat | 77.67 | #507 |
| 3 | AIC-1 | 77.67 | #231 |
| 4 | free-evo-qwen72B-v0.8-re | 76.7 | #435 |
| 5 | AceGPT-v2-70B-Chat | 75.73 | #407 |
| 6 | Qwen 2 72B | 74.76 | #521 |
| 7 | Llama 3.3 70B Instruct | 72.82 | #534 |
| 8 | Qwen 2.5 14B | 72.82 | #725 |
| 9 | Gemma 3 27B (IT) | 70.87 | #523 |
| 10 | Qwentile2.5-32B-Instruct | 70.87 | #339 |
| 11 | tempmotacilla-cinerea-0308 | 70.87 | #395 |
| 12 | Hala-9B | 70.87 | #504 |
| 13 | Qwen2.5-Lumen-14B | 70.87 | #477 |
| 14 | aya-expanse-8B | 69.9 | #931 |
| 15 | lambda-qwen2.5-14B-dpo-test | 69.9 | #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-ht-management · How It Works · Data refreshed daily, snapshot 2026-10-20.