Open Arabic LLM - Arabic MMLU HT College Physics: leaderboard
The "Arabic MMLU HT College Physics" 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 July 2027. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 27B (IT) | 55.88 | #523 |
| 2 | Qwen 2.5 72B Instruct | 54.9 | #454 |
| 3 | AceGPT-v2-70B-Chat | 54.9 | #407 |
| 4 | gemma-3-12B-pt | 53.92 | #686 |
| 5 | Qwen3-8B-Base | 52.94 | #626 |
| 6 | Qwen2.5-32B-Instruct-abliterated-v2 | 50.98 | #394 |
| 7 | gemma-3-27B-pt | 50 | #502 |
| 8 | Qwen2.5-32B-Instruct-CFT | 49.02 | #375 |
| 9 | Hala-9B | 49.02 | #504 |
| 10 | Qwen 2.5 32B Instruct | 48.04 | #498 |
| 11 | tempmotacilla-cinerea-0308 | 48.04 | #395 |
| 12 | Qwen 2.5 14B | 47.06 | #725 |
| 13 | Qwen 2.5 72B | 46.08 | #499 |
| 14 | Saka-14B | 46.08 | #344 |
| 15 | Qwen3-4B-Base | 45.1 | #897 |
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-mmlu-ht-college-physics · How It Works · Data refreshed daily, snapshot 2026-10-20.