Open Arabic LLM - Aratrust Offensive: leaderboard
The "Aratrust Offensive" 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 | Gemma 3 27B (IT) | 98.55 | #523 |
| 2 | Gemma 2 27B (IT) | 98.55 | #645 |
| 3 | Hala-9B | 97.1 | #504 |
| 4 | gemma-3-4B-pt Base | 95.65 | #1134 |
| 5 | gemma-3-27B-pt | 95.65 | #502 |
| 6 | tempmotacilla-cinerea-0308 | 95.65 | #395 |
| 7 | Gemma 2 9B (IT) | 94.2 | #788 |
| 8 | aya-23-35B | 94.2 | #883 |
| 9 | ALLaM-7B-Instruct-preview | 94.2 | #718 |
| 10 | jais-adapted-13B-chat | 94.2 | #957 |
| 11 | jais-family-30B-16k-chat | 94.2 | #755 |
| 12 | EZO-Qwen2.5-32B-Instruct | 94.2 | #471 |
| 13 | gemma-3-12B-pt | 92.75 | #686 |
| 14 | Fanar-1-9B-Instruct | 92.75 | #633 |
| 15 | Bielik-11B-v2 | 92.75 | #1329 |
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-offensive · How It Works · Data refreshed daily, snapshot 2026-10-20.