GroundedSurg: leaderboard
Metric: Mask IoU (0-1, times 100) between the projected mask and the ground-truth instrument mask, averaged over instances, on GroundedSurg (about 1,071 language-referenced surgical instrument instances in 612 images from four procedures), zero-shot: the model predicts a bounding box and centre point for the referred instrument and a frozen SAM3 backend turns them into a mask; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 20 | #643 |
| 2 | GPT-5.2 | 9 | #105 |
| 3 | Qwen 3 VL 8B Instruct | 9 | #401 |
| 4 | Gemma 3 12B | 7 | #666 |
| 5 | Gemma 3 27B | 6 | #596 |
| 6 | GPT-4o Mini | 4 | #588 |
| 7 | Ministral 3 8B | 3 | #676 |
| 8 | MedGemma-4B-IT | 2 | #842 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=groundedsurg · How It Works · Data refreshed daily, snapshot 2026-10-11.