Open Arabic LLM - Arabic Exams: leaderboard
The "Arabic Exams" 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: Around January 2027. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 66.67 | #534 |
| 2 | c4ai-command-r7B-arabic-02-2025 | 64.99 | #732 |
| 3 | Qwen 2.5 72B Instruct | 58.66 | #454 |
| 4 | Gemma 3 27B (IT) | 57.54 | #523 |
| 5 | gemma-3-27B-pt | 57.36 | #502 |
| 6 | AIC-1 | 56.61 | #231 |
| 7 | Qwen2.5-32B-Instruct-abliterated-v2 | 55.87 | #394 |
| 8 | Awqward2.5-32B-Instruct | 55.87 | #371 |
| 9 | Saka-14B | 55.68 | #344 |
| 10 | Qwen2.5-32B-Instruct-CFT | 55.68 | #375 |
| 11 | lambda-qwen2.5-14B-dpo-test | 55.49 | #481 |
| 12 | AceGPT-v2-70B-Chat | 55.31 | #407 |
| 13 | Qwen 2.5 32B Instruct | 55.12 | #498 |
| 14 | gemma-3-12B-pt | 54.38 | #686 |
| 15 | EZO-Qwen2.5-32B-Instruct | 54.38 | #471 |
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-arabic-exams · How It Works · Data refreshed daily, snapshot 2026-10-20.