Open Arabic LLM - Arabic MMLU Social Science (Middle School): leaderboard
The "Arabic MMLU Social Science (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: Around April 2027. 163 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | c4ai-command-r7B-arabic-02-2025 | 63.49 | #732 |
| 2 | gemma-3-27B-pt | 60.58 | #502 |
| 3 | Saka-14B | 60.17 | #344 |
| 4 | Qwen 2.5 14B | 58.51 | #725 |
| 5 | lambda-qwen2.5-14B-dpo-test | 58.09 | #481 |
| 6 | tempmotacilla-cinerea-0308 | 58.09 | #395 |
| 7 | Qwen 2.5 72B Instruct | 57.68 | #454 |
| 8 | Qwen 2.5 32B Instruct | 57.68 | #498 |
| 9 | free-evo-qwen72B-v0.8-re | 57.68 | #435 |
| 10 | Llama 3.3 70B Instruct | 57.26 | #534 |
| 11 | jais-adapted-70B-chat | 56.85 | #636 |
| 12 | Awqward2.5-32B-Instruct | 56.85 | #371 |
| 13 | Qwen2.5-32B-Instruct-CFT | 56.85 | #375 |
| 14 | AIC-1 | 56.85 | #231 |
| 15 | Qwentile2.5-32B-Instruct | 56.43 | #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-social-science-middle-school · How It Works · Data refreshed daily, snapshot 2026-10-20.