SpatialEdit-Bench - Object Moving: leaderboard
Metric: Moving score (0-1): square root of the detector IoU with the target box times a VLM object-consistency score, for instruction-based object translation and scaling; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | LongCat-Image-Edit | 0.37 |
| 2 | Qwen-Image-Edit | 0.31 |
| 3 | Edit-R1 (SpatialEdit-Bench checkpoint unspecified) | 0.31 |
| 4 | Seedream 4.0 | 0.16 |
| 5 | Nano Banana | 0.1 |
Interactive version: theaggregate.ai/benchmark?slug=spatialedit-bench-object-moving · How It Works · Data refreshed daily, snapshot 2026-10-07.