Open Arabic LLM - Aratrust Average: leaderboard
The "Aratrust 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: Estimated already saturated. 162 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | tempmotacilla-cinerea-0308 | 89.98 | #395 |
| 2 | Hala-9B | 89.98 | #504 |
| 3 | AIC-1 | 89.83 | #231 |
| 4 | Qwen 2 72B | 89.22 | #521 |
| 5 | Qwen 2.5 72B Instruct | 89.21 | #454 |
| 6 | EZO-Qwen2.5-32B-Instruct | 89.15 | #471 |
| 7 | Gemma 3 27B (IT) | 89.08 | #523 |
| 8 | aya-expanse-32B | 89 | #692 |
| 9 | lambda-qwen2.5-14B-dpo-test | 88.83 | #481 |
| 10 | Qwen 2.5 72B | 88.55 | #499 |
| 11 | Qwen2.5-32B-Instruct-abliterated-v2 | 88.4 | #394 |
| 12 | Fanar-1-9B-Instruct | 88.29 | #633 |
| 13 | Saka-14B | 88.29 | #344 |
| 14 | Llama 3.3 70B Instruct | 88.27 | #534 |
| 15 | Awqward2.5-32B-Instruct | 88.27 | #371 |
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-aratrust-average · How It Works · Data refreshed daily, snapshot 2026-10-20.