RE-Edit - Environmental: leaderboard
RE-Edit (2026) tests instruction-based image editors on 1,000 curated requests whose correct edit depends on implicit reasoning. This board is the pass rate on the environmental-reasoning subset (scene and environmental context), with a Qwen3-VL-30B judge marking each edited image pass or fail against the case rationale. Higher is better.
Metric: Pass rate (%) on the environmental-reasoning subset of RE-Edit's 1,000 curated edit requests (scene and environmental context): a Qwen3-VL-30B judge marks each edited image pass or fail against the case rationale; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seedream 4.0 | 16.3 | |
| 2 | FLUX.2-dev | 14.8 | |
| 3 | Ovis-U1-3B | 13.3 | #2318 |
| 4 | Qwen-Image-Edit | 12.8 | |
| 5 | Nano Banana | 10.8 | |
| 6 | Step1X-Edit-v1p2-preview | 9.4 | |
| 7 | DreamOmni2 | 9.4 | |
| 8 | HiDream-E1 | 8.9 | |
| 9 | Step1X-Edit-v1p1 | 7.4 | |
| 10 | FLUX.1 Kontext [dev] | 4.4 |
Interactive version: theaggregate.ai/benchmark?slug=re-edit-environmental · How It Works · Data refreshed daily, snapshot 2026-11-13.