Open Arabic LLM - Arabic MMLU HT Logical Fallacies: leaderboard
The "Arabic MMLU HT Logical Fallacies" 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 | 76.69 | #454 |
| 2 | Qwen 2 72B | 73.01 | #521 |
| 3 | Gemma 3 27B (IT) | 71.17 | #523 |
| 4 | Qwentile2.5-32B-Instruct | 71.17 | #339 |
| 5 | gemma-3-27B-pt | 70.55 | #502 |
| 6 | Hala-9B | 69.94 | #504 |
| 7 | Awqward2.5-32B-Instruct | 69.33 | #371 |
| 8 | Llama 3.3 70B Instruct | 68.71 | #534 |
| 9 | Qwen 2.5 72B | 68.71 | #499 |
| 10 | Qwen2.5-32B-Instruct-abliterated-v2 | 68.71 | #394 |
| 11 | AIC-1 | 68.71 | #231 |
| 12 | Gemma 2 27B (IT) | 68.1 | #645 |
| 13 | jais-adapted-70B-chat | 67.48 | #636 |
| 14 | Qwen2.5-32B-Instruct-CFT | 67.48 | #375 |
| 15 | Qwen 2.5 32B Instruct | 66.87 | #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-ht-logical-fallacies · How It Works · Data refreshed daily, snapshot 2026-10-20.