PaintBench - Color Change: leaderboard
PaintBench mIoU (%) averaged over its five color change tasks (recoloring, flood fill, blending, gradients and point operations); edits are scored pixel by pixel against the single correct image, higher is better (NYU, 2026).
Metric: mIoU (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nano-Banana-2 | 17.2 | |
| 2 | GPT-Image-2 | 13.8 | |
| 3 | Nano Banana | 6.4 | |
| 4 | Qwen-Image-Edit-2511 | 5.4 | |
| 5 | FLUX.2-dev | 2.7 | |
| 6 | BAGEL-7B-MoT | 2.6 | #5515 |
| 7 | FLUX.2-klein-9B | 2.4 | |
| 8 | LongCat-Image-Edit | 1.8 | |
| 9 | FLUX.1-Kontext-dev | 1.6 | |
| 10 | HunyuanImage-3.0 | 0.2 |
Interactive version: theaggregate.ai/benchmark?slug=paintbench-color-change · How It Works · Data refreshed daily, snapshot 2026-11-12.