Open Arabic LLM - Aratrust Unfairness: leaderboard
The "Aratrust Unfairness" 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 | 96.36 | #499 |
| 2 | lambda-qwen2.5-14B-dpo-test | 96.36 | #481 |
| 3 | Hala-9B | 96.36 | #504 |
| 4 | Qwen 2.5 72B Instruct | 94.55 | #454 |
| 5 | Qwen 2.5 14B Instruct | 94.55 | #647 |
| 6 | Phi-4 Mini Instruct | 94.55 | #1245 |
| 7 | Qwen 2 72B | 94.55 | #521 |
| 8 | EZO-Qwen2.5-32B-Instruct | 94.55 | #471 |
| 9 | Qwen2.5-14B-Gutenberg-1e-Delta | 94.55 | #473 |
| 10 | Rombos-LLM-V2.6-Qwen-14B | 94.55 | #516 |
| 11 | recoilme-gemma-2-9B-v0.2 | 94.55 | #629 |
| 12 | Qwen2.5-Lumen-14B | 94.55 | #477 |
| 13 | AIC-1 | 94.55 | #231 |
| 14 | gemma-3-12B-pt | 92.73 | #686 |
| 15 | aya-expanse-32B | 92.73 | #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-unfairness · How It Works · Data refreshed daily, snapshot 2026-10-20.