UniEditBench (Image): leaderboard
Metric: Mean (1-5) of the four dimension scores (structural fidelity, text alignment, background consistency, naturalness); 633 image-editing samples over nine operations, each edit scored on a 1-5 Likert scale by the authors' Qwen3-VL-8B evaluator distilled from Qwen3-VL-235B-A22B (the second entry of each 4B/8B cell); higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | UniEditBench Hunyuan (checkpoint unspecified) | 4.4 |
| 2 | FLUX.1-Kontext-dev | 4.3 |
| 3 | Qwen-Image-Edit | 4.29 |
| 4 | Step1X-Edit | 4.21 |
| 5 | DreamOmni2 | 4.1 |
| 6 | UniWorld-V2 | 4.02 |
| 7 | OmniGen2 | 3.99 |
| 8 | BAGEL-7B-MoT | 3.96 |
| 9 | BAGEL-7B-MoT (Thinking) | 3.91 |
| 10 | VAREdit | 3.79 |
Interactive version: theaggregate.ai/benchmark?slug=unieditbench-image · How It Works · Data refreshed daily, snapshot 2026-10-07.