Open Arabic LLM - Arabic MMLU HT High School Mathematics: leaderboard
The "Arabic MMLU HT High School Mathematics" 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 2031. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | AIC-1 | 52.96 | #231 |
| 2 | Saka-14B | 49.63 | #344 |
| 3 | gemma-3-27B-pt | 47.78 | #502 |
| 4 | Qwen3-8B-Base | 46.67 | #626 |
| 5 | Qwen 2.5 14B | 45.19 | #725 |
| 6 | Qwen 2 72B | 44.07 | #521 |
| 7 | Qwen2.5-32B-Instruct-abliterated-v2 | 42.96 | #394 |
| 8 | AceGPT-v2-32B-Chat | 41.85 | #507 |
| 9 | AceGPT-v2-70B-Chat | 41.85 | #407 |
| 10 | Qwen 2.5 7B | 41.48 | #922 |
| 11 | tempmotacilla-cinerea-0308 | 41.11 | #395 |
| 12 | Qwen 2.5 72B Instruct | 40.74 | #454 |
| 13 | Qwen3-4B-Base | 40.37 | #897 |
| 14 | Llama 3.3 70B Instruct | 40 | #534 |
| 15 | Qwen 2 7B Instruct | 39.63 | #991 |
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-high-school-mathematics · How It Works · Data refreshed daily, snapshot 2026-10-20.