Open Arabic LLM - Alghafa Meta AR Dialects: leaderboard
The "Alghafa Meta AR Dialects" 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. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 83.71 | #454 |
| 2 | AceGPT-v2-70B-Chat | 82.74 | #407 |
| 3 | Awqward2.5-32B-Instruct | 82.71 | #371 |
| 4 | Qwen 2.5 32B Instruct | 82.54 | #498 |
| 5 | Qwen2.5-32B-Instruct-CFT | 82.47 | #375 |
| 6 | AIC-1 | 82.39 | #231 |
| 7 | Qwentile2.5-32B-Instruct | 82.21 | #339 |
| 8 | Qwen2.5-32B-Instruct-abliterated-v2 | 81.63 | #394 |
| 9 | Gemma 3 27B (IT) | 81.26 | #523 |
| 10 | Qwen 2.5 32B | 79.87 | #546 |
| 11 | gemma-3-27B-pt | 79.31 | #502 |
| 12 | Saka-14B | 79.28 | #344 |
| 13 | AceGPT-v2-32B-Chat | 78.63 | #507 |
| 14 | Qwen 2.5 72B | 78.35 | #499 |
| 15 | Qwen 2 72B | 78.13 | #521 |
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-alghafa-meta-ar-dialects · How It Works · Data refreshed daily, snapshot 2026-10-20.