QIMMA - MedArabiQ QA: leaderboard
Open-ended question-answering part of MedArabiQ, Arabic medical questions, as run in QIMMA. Generated answers are scored by BERTScore F1 against references (0-100, higher is better); BERTScore has a high floor, so models span a narrow range.
Metric: BERTScore F1 (%). Source: huggingface.co. Saturation forecast: Around 2031. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 26B A4B (IT) | 56.88 | #293 |
| 2 | Gemma 4 31B (IT) | 56.73 | #178 |
| 3 | Qwen 3 235B A22B 2507 Instruct | 55.59 | #315 |
| 4 | Gemma 3 27B (IT) | 55.45 | #528 |
| 5 | AIC-1 | 54.95 | #235 |
| 6 | Jais-2-70B-Chat | 54.67 | #430 |
| 7 | Qwen 2.5 72B Instruct | 54.54 | #469 |
| 8 | Llama 3.3 70B Instruct | 54.29 | #541 |
| 9 | Qwen 2.5 32B Instruct | 54.28 | #509 |
| 10 | Fanar-1-9B-Instruct | 54.14 | #635 |
| 11 | AceGPT-v2-32B-Chat | 54.04 | #518 |
| 12 | gemma-4-E4B-it | 53.78 | #577 |
| 13 | LFM2 24B A2B | 53.6 | #1362 |
| 14 | Qwen 3 30B A3B 2507 Instruct | 53.58 | #492 |
| 15 | ALLaM-7B-Instruct-preview | 53.55 | #724 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qimma-medarabiq-qa · How It Works · Data refreshed daily, snapshot 2026-11-05.