Open Arabic LLM - Arabic MMLU Physics (High School): leaderboard
The "Arabic MMLU Physics (High School)" 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 | gemma-3-27B-pt | 62.35 | #502 |
| 2 | Qwen3-8B-Base | 61.96 | #626 |
| 3 | Llama 3.3 70B Instruct | 61.18 | #534 |
| 4 | Qwen2.5-32B-Instruct-abliterated-v2 | 60.39 | #394 |
| 5 | Qwen 2.5 72B Instruct | 60 | #454 |
| 6 | AIC-1 | 59.61 | #231 |
| 7 | AceGPT-v2-32B-Chat | 58.04 | #507 |
| 8 | Qwentile2.5-32B-Instruct | 56.86 | #339 |
| 9 | Awqward2.5-32B-Instruct | 56.86 | #371 |
| 10 | Saka-14B | 56.08 | #344 |
| 11 | Qwen 2.5 32B Instruct | 55.69 | #498 |
| 12 | Qwen2.5-32B-Instruct-CFT | 55.69 | #375 |
| 13 | Qwen 2.5 14B | 54.9 | #725 |
| 14 | AceGPT-v2-70B-Chat | 54.9 | #407 |
| 15 | Qwen 2 72B | 54.51 | #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-arabic-mmlu-physics-high-school · How It Works · Data refreshed daily, snapshot 2026-10-20.