Colon-Bench - Lesion Segmentation mIoU: leaderboard
Metric: Mean IoU (%) of lesion masks over all evaluated frames of Colon-Bench's segmentation clips: the model boxes the named lesion on three frames and a fixed EdgeTAM tracker propagates the masks; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 45 | #77 |
| 2 | Gemini 3.1 Flash Lite | 37.4 | #174 |
| 3 | GPT-5.4 | 34.5 | #76 |
| 4 | GPT-5.2 | 30.7 | #105 |
| 5 | Gemini 2.5 Flash Lite | 19.9 | #413 |
| 6 | Qwen 3.5 Plus | 16.7 | #123 |
| 7 | Qwen 3.5 397B A17B | 16.6 | #141 |
| 8 | Claude Opus 4.6 | 16.1 | #60 |
| 9 | Qwen 3 VL 235B A22B | 13.6 | #228 |
| 10 | Seed-1.6 | 12.6 | #257 |
| 11 | GLM-4.6V | 12.5 | #309 |
| 12 | Molmo2-8B | 2.6 | #522 |
| 13 | Seed 1.6 Flash | 2.6 | #427 |
| 14 | GPT-4o | 0.5 | #333 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=colon-bench-lesion-segmentation-miou · How It Works · Data refreshed daily, snapshot 2026-10-11.