Open Arabic LLM - Aratrust Privacy: leaderboard
The "Aratrust Privacy" 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 | Hala-9B | 100 | #504 |
| 2 | Llama 3.3 70B Instruct | 98.25 | #534 |
| 3 | Qwen 2 72B | 98.25 | #521 |
| 4 | lambda-qwen2.5-14B-dpo-test | 98.25 | #481 |
| 5 | recoilme-gemma-2-9B-v0.4 | 98.25 | #553 |
| 6 | Qwen 2.5 14B Instruct | 96.49 | #647 |
| 7 | Qwen 2.5 72B | 96.49 | #499 |
| 8 | AceGPT-v2-32B-Chat | 96.49 | #507 |
| 9 | Qwen2.5-32B-Instruct-abliterated-v2 | 96.49 | #394 |
| 10 | Qwentile2.5-32B-Instruct | 96.49 | #339 |
| 11 | EZO-Qwen2.5-32B-Instruct | 96.49 | #471 |
| 12 | Qwen2.5-14B-Gutenberg-1e-Delta | 96.49 | #473 |
| 13 | Rombos-LLM-V2.6-Qwen-14B | 96.49 | #516 |
| 14 | free-evo-qwen72B-v0.8-re | 96.49 | #435 |
| 15 | Qwen2.5-Lumen-14B | 96.49 | #477 |
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-privacy · How It Works · Data refreshed daily, snapshot 2026-10-20.