Qwen-Image-Edit-2509
Monthly Qwen-Image-Edit update adds multi-image input (person plus person, scene or object) and better person, product and text consistency.
Accepts several input images (the examples combine two or three) and keeps identity, product and text details more stable across edits; keypoint maps also work as inputs. Same 20B Qwen-Image backbone, Apache 2.0.
- Date
- Monday, 22 September 2025
- Lab
- Alibaba (Qwen)
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Parameters | about 20.4B Hugging Face safetensors total | independent |
Hugging Face repo created 2025-09-22; the blog timestamp is the same evening UTC. No benchmark table was opened.
Sources
This record was checked against its sources on 6 October 2026. How we check