Open Arabic LLM - Arabic MMLU HT Prehistory: leaderboard
The "Arabic MMLU HT Prehistory" 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 | Qwen 2.5 72B Instruct | 75.93 | #454 |
| 2 | Llama 3.3 70B Instruct | 72.84 | #534 |
| 3 | Qwen 2.5 72B | 72.53 | #499 |
| 4 | AceGPT-v2-70B-Chat | 70.99 | #407 |
| 5 | Qwentile2.5-32B-Instruct | 70.68 | #339 |
| 6 | AIC-1 | 70.06 | #231 |
| 7 | Awqward2.5-32B-Instruct | 69.14 | #371 |
| 8 | Gemma 3 27B (IT) | 68.52 | #523 |
| 9 | Qwen 2 72B | 68.21 | #521 |
| 10 | Qwen2.5-32B-Instruct-CFT | 68.21 | #375 |
| 11 | Qwen 2.5 32B Instruct | 67.9 | #498 |
| 12 | gemma-3-27B-pt | 67.9 | #502 |
| 13 | Qwen2.5-32B-Instruct-abliterated-v2 | 67.28 | #394 |
| 14 | AceGPT-v2-32B-Chat | 65.43 | #507 |
| 15 | gemma-3-12B-pt | 64.81 | #686 |
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-prehistory · How It Works · Data refreshed daily, snapshot 2026-10-20.