QIMMA - MedAraBench: leaderboard
MedAraBench Arabic medical multiple-choice questions, as run in the QIMMA Arabic leaderboard. Scored with LightEval log-likelihood over the answer options with character-length normalization rather than generated answers, so reasoning-tuned models can land near chance; accuracy 0-100, higher is better.
Metric: Accuracy (%, log-likelihood multiple choice). Source: huggingface.co. Saturation forecast: Not forecast. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 54.86 | #541 |
| 2 | Gemma 4 31B | 54.5 | #212 |
| 3 | Qwen 3.5 122B A10B | 52.49 | #184 |
| 4 | Jais-2-70B-Chat | 50.89 | #430 |
| 5 | AIC-1 | 50.78 | #235 |
| 6 | Fanar-1-9B-Instruct | 48.53 | #635 |
| 7 | Qwen 3.5 397B A17B | 47.97 | #154 |
| 8 | Gemma 4 26B A4B | 47.77 | #360 |
| 9 | Hala-9B | 47.54 | #517 |
| 10 | AceGPT-v2-32B-Chat | 47.32 | #518 |
| 11 | Qwen 3.5 35B A3B | 46.46 | #268 |
| 12 | Qwen 3.5 35B A3B Base | 46.32 | #319 |
| 13 | Qwen 2.5 72B Instruct | 44.19 | #469 |
| 14 | gemma-3-27B-pt | 43.72 | #505 |
| 15 | Atlas-Chat-9B | 42.52 | #710 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qimma-medarabench · How It Works · Data refreshed daily, snapshot 2026-11-05.