QIMMA - ArabicMMLU: leaderboard
Metric: Accuracy (%, log-likelihood multiple choice). Source: huggingface.co. 83 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Jais-2-70B-Chat | 81.29 |
| 2 | Qwen 3.5 397B A17B | 77.54 |
| 3 | Gemma 4 31B | 77.25 |
| 4 | Jais-2-8B-Chat | 73.96 |
| 5 | Qwen 2.5 72B Instruct | 73.78 |
| 6 | Qwen 3.5 122B A10B | 73.17 |
| 7 | ALLaM-7B-Instruct-preview | 72.54 |
| 8 | gemma-3-27B-pt | 72.27 |
| 9 | AIC-1 | 72.02 |
| 10 | Llama 3.3 70B Instruct | 71.57 |
| 11 | Gemma 4 26B A4B | 71.14 |
| 12 | AceGPT-v2-32B-Chat | 70.62 |
| 13 | Qwen 3.5 27B | 69.47 |
| 14 | Qwen 2.5 32B Instruct | 68.76 |
| 15 | Fanar-1-9B-Instruct | 67.39 |
Interactive version: theaggregate.ai/benchmark?slug=qimma-arabicmmlu · How It Works · Data refreshed daily, snapshot 2026-09-19.