Open Arabic LLM - Aratrust Ethics: leaderboard
The "Aratrust Ethics" 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 | ALLaM-7B-Instruct-preview | 83.33 | #718 |
| 2 | Yehia-7B-preview | 83.33 | #646 |
| 3 | Qwen2.5-32B-Instruct-abliterated-v2 | 81.67 | #394 |
| 4 | Awqward2.5-32B-Instruct | 81.67 | #371 |
| 5 | tempmotacilla-cinerea-0308 | 81.67 | #395 |
| 6 | Qwen2.5-32B-Instruct-CFT | 81.67 | #375 |
| 7 | Qwen 2.5 72B Instruct | 80 | #454 |
| 8 | Qwen 2.5 32B Instruct | 80 | #498 |
| 9 | gemma-3-12B-pt | 80 | #686 |
| 10 | Qwen 2.5 72B | 80 | #499 |
| 11 | Qwen 2 72B | 80 | #521 |
| 12 | Qwentile2.5-32B-Instruct | 80 | #339 |
| 13 | EZO-Qwen2.5-32B-Instruct | 80 | #471 |
| 14 | Llama 3.3 70B Instruct | 78.33 | #534 |
| 15 | aya-expanse-32B | 78.33 | #692 |
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-ethics · How It Works · Data refreshed daily, snapshot 2026-10-20.