Open Arabic LLM - Aratrust Trustfulness: leaderboard
The "Aratrust Trustfulness" 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 | Llama 3.3 70B Instruct | 82.05 | #534 |
| 2 | ALLaM-7B-Instruct-preview | 82.05 | #718 |
| 3 | Yehia-7B-preview | 82.05 | #646 |
| 4 | AIC-1 | 82.05 | #231 |
| 5 | Gemma 3 27B (IT) | 80.77 | #523 |
| 6 | Qwen 2.5 32B Instruct | 80.77 | #498 |
| 7 | Qwen 2.5 72B Instruct | 79.49 | #454 |
| 8 | aya-expanse-32B | 79.49 | #692 |
| 9 | gemma-3-27B-pt | 79.49 | #502 |
| 10 | Qwen 2.5 72B | 78.21 | #499 |
| 11 | AceGPT-v2-70B-Chat | 78.21 | #407 |
| 12 | Qwen2.5-32B-Instruct-abliterated-v2 | 78.21 | #394 |
| 13 | Awqward2.5-32B-Instruct | 78.21 | #371 |
| 14 | EZO-Qwen2.5-32B-Instruct | 76.92 | #471 |
| 15 | Saka-14B | 76.92 | #344 |
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-trustfulness · How It Works · Data refreshed daily, snapshot 2026-10-20.