Open Arabic LLM - Madinah QA Arabic Language (Grammar): leaderboard
The "Madinah QA Arabic Language (Grammar)" 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 March 2027. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | AIC-1 | 74.79 | #231 |
| 2 | free-evo-qwen72B-v0.8-re | 70.96 | #435 |
| 3 | AceGPT-v2-32B-Chat | 69.59 | #507 |
| 4 | Fanar-1-9B-Instruct | 68.77 | #633 |
| 5 | Qwentile2.5-32B-Instruct | 67.95 | #339 |
| 6 | Hala-9B | 67.12 | #504 |
| 7 | recoilme-gemma-2-9B-v0.4 | 63.29 | #553 |
| 8 | tempmotacilla-cinerea-0308 | 62.47 | #395 |
| 9 | Saka-14B | 61.1 | #344 |
| 10 | Qwen2.5-7B-Instruct-abliterated | 60.82 | #790 |
| 11 | Qwen 2.5 7B Instruct | 59.45 | #862 |
| 12 | Qwen2.5-7B-Instruct-Uncensored | 59.18 | #684 |
| 13 | Gemma 3 27B (IT) | 58.63 | #523 |
| 14 | jais-family-30B-16k-chat | 56.71 | #755 |
| 15 | AceGPT-v2-70B-Chat | 55.62 | #407 |
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-madinah-qa-arabic-language-grammar · How It Works · Data refreshed daily, snapshot 2026-10-20.