Open Arabic LLM - Arabic MMLU Social Science (Primary School): leaderboard
The "Arabic MMLU Social Science (Primary 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 | Saka-14B | 82.27 | #344 |
| 2 | Qwen 2.5 72B Instruct | 81.28 | #454 |
| 3 | gemma-3-27B-pt | 80.43 | #502 |
| 4 | lambda-qwen2.5-14B-dpo-test | 80 | #481 |
| 5 | Llama 3.3 70B Instruct | 79.72 | #534 |
| 6 | tempmotacilla-cinerea-0308 | 79.72 | #395 |
| 7 | Qwen 2.5 14B | 79.57 | #725 |
| 8 | Qwen 2.5 72B | 79.43 | #499 |
| 9 | AIC-1 | 78.72 | #231 |
| 10 | Qwen2.5-32B-Instruct-abliterated-v2 | 78.44 | #394 |
| 11 | Fanar-1-9B-Instruct | 78.3 | #633 |
| 12 | Awqward2.5-32B-Instruct | 78.3 | #371 |
| 13 | Gemma 3 27B (IT) | 78.16 | #523 |
| 14 | AceGPT-v2-32B-Chat | 78.16 | #507 |
| 15 | Qwen 2.5 32B Instruct | 77.87 | #498 |
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-primary-school · How It Works · Data refreshed daily, snapshot 2026-10-20.