Lingshu-7B: benchmark results
lingshu-medical-mllm's Lingshu 7B, a multimodal medical vision-language model supporting over a dozen imaging modalities including X-ray, CT and MRI. Provider: Alibaba. Released 2025-06-05. Access: Open.
Aggregate Rating 1557 ± 13, rank #1599 of 10916 rated models, from 214 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PathBind - Text-only PTA D5 | 46.7 | Score (%) | 100 |
| CardioLens | 56.22 | F1 (Random) (self-reported) | 95.7 |
| Medical MM Leaderboard - PathVQA | 61.9 | Accuracy (%) | 95.2 |
| Medical MM Leaderboard - SLAKE | 83.1 | Accuracy (%) | 95.2 |
| Medical MM Leaderboard - OMVQA | 82.9 | Accuracy (%) | 94.7 |
| EviPathBench - Diagnostic Reasoning | 93.41 | Accuracy (%) pooled over all questions (the All part of each | 94.4 |
| PathBind - Image PTA D3 | 86.7 | Score (%) | 94.1 |
| PathBind - Text-only PTA D1 | 44.2 | Score (%) | 94.1 |
| Med-R2 - Image Quality | 67.3 | Accuracy (%) on the Med-R2 Image Quality task (judging the q | 92.3 |
| PathBind - Text-only PTA D3 | 65 | Score (%) | 91.2 |
| MM-NeuroOnco (Closed-Ended) | 37.6 | Accuracy (%) on all closed-ended questions; five-option mult | 88.9 |
| MM-NeuroOnco (Closed-Ended) - Diagnosis | 39.2 | Accuracy (%) on tumor-type diagnosis; five-option multiple c | 88.9 |
Interactive version: theaggregate.ai/model?slug=lingshu-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.