Open Arabic LLM - Aratrust Illegal: leaderboard
The "Aratrust Illegal" 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 | Qwen 2.5 72B Instruct | 94.34 | #454 |
| 2 | aya-expanse-32B | 94.34 | #692 |
| 3 | Qwen3-8B-Base | 94.34 | #626 |
| 4 | AceGPT-v2-70B-Chat | 94.34 | #407 |
| 5 | Qwen 2 72B | 94.34 | #521 |
| 6 | tempmotacilla-cinerea-0308 | 94.34 | #395 |
| 7 | Hala-9B | 94.34 | #504 |
| 8 | AIC-1 | 94.34 | #231 |
| 9 | AceGPT-v2-32B | 94.34 | #768 |
| 10 | Gemma 3 27B (IT) | 92.45 | #523 |
| 11 | Qwen 2.5 7B Instruct | 92.45 | #862 |
| 12 | aya-23-35B | 92.45 | #883 |
| 13 | Qwen 2.5 72B | 92.45 | #499 |
| 14 | Qwen 2.5 14B | 92.45 | #725 |
| 15 | Qwen3-4B-Base | 92.45 | #897 |
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-illegal · How It Works · Data refreshed daily, snapshot 2026-10-20.