Open Arabic LLM - Arabic MMLU HT Public Relations: leaderboard
The "Arabic MMLU HT Public Relations" 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 (IT) | 68.18 | #523 |
| 2 | AceGPT-v2-70B-Chat | 66.36 | #407 |
| 3 | Qwen 2 72B | 66.36 | #521 |
| 4 | Qwentile2.5-32B-Instruct | 66.36 | #339 |
| 5 | Qwen 2.5 72B Instruct | 65.45 | #454 |
| 6 | free-evo-qwen72B-v0.8-re | 65.45 | #435 |
| 7 | aya-expanse-32B | 64.55 | #692 |
| 8 | gemma-3-27B-pt | 64.55 | #502 |
| 9 | AceGPT-v2-32B-Chat | 63.64 | #507 |
| 10 | jais-family-30B-16k-chat | 63.64 | #755 |
| 11 | Fanar-1-9B-Instruct | 63.64 | #633 |
| 12 | recoilme-gemma-2-9B-v0.4 | 63.64 | #553 |
| 13 | Awqward2.5-32B-Instruct | 62.73 | #371 |
| 14 | AceGPT-v2-32B | 62.73 | #768 |
| 15 | Qwen 2.5 32B Instruct | 61.82 | #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-ht-public-relations · How It Works · Data refreshed daily, snapshot 2026-10-20.