QIMMA - MedArabiQ QA: leaderboard

Metric: BERTScore F1 (%). Source: huggingface.co. 83 models tracked.

Top models

#ModelScore
1Gemma 4 26B A4B (IT)56.88
2Gemma 4 31B (IT)56.73
3Qwen 3 235B A22B 2507 Instruct55.59
4Gemma 3 27B (IT)55.45
5AIC-154.95
6Jais-2-70B-Chat54.67
7Qwen 2.5 72B Instruct54.54
8Llama 3.3 70B Instruct54.29
9Qwen 2.5 32B Instruct54.28
10Fanar-1-9B-Instruct54.14
11AceGPT-v2-32B-Chat54.04
12gemma-4-E4B-it53.78
13LFM2 24B A2B53.6
14Qwen 3 30B A3B 2507 Instruct53.58
15ALLaM-7B-Instruct-preview53.55

Interactive version: theaggregate.ai/benchmark?slug=qimma-medarabiq-qa · How It Works · Data refreshed daily, snapshot 2026-09-19.