Open Arabic LLM - Arabic MMLU History (Middle School): leaderboard
The "Arabic MMLU History (Middle 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: Estimated already saturated. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | jais-adapted-70B-chat | 73.89 | #636 |
| 2 | ALLaM-7B-Instruct-preview | 73.4 | #718 |
| 3 | Llama 3.3 70B Instruct | 72.91 | #534 |
| 4 | Qwen 2.5 72B Instruct | 72.41 | #454 |
| 5 | Qwen 2.5 14B | 72.41 | #725 |
| 6 | Yehia-7B-preview | 72.41 | #646 |
| 7 | gemma-3-27B-pt | 71.92 | #502 |
| 8 | Saka-14B | 70.94 | #344 |
| 9 | AceGPT-v2-32B-Chat | 70.44 | #507 |
| 10 | aya-23-35B | 69.46 | #883 |
| 11 | tempmotacilla-cinerea-0308 | 69.46 | #395 |
| 12 | gemma-3-12B-pt | 68.47 | #686 |
| 13 | Hala-9B | 68.47 | #504 |
| 14 | jais-adapted-13B-chat | 67.98 | #957 |
| 15 | Qwentile2.5-32B-Instruct | 67.98 | #339 |
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-history-middle-school · How It Works · Data refreshed daily, snapshot 2026-10-20.