# Qwen open-weights Qwen-Image-2.1, a 7B unified image generation and editing model

Qwen announced Qwen-Image-2.1, saying the 7B model unifies image generation and editing and that its weights are now open. The company said it natively handles RGBA layers, supports up to 10 reference images, and offers precise local control.

Language: en
Time zone: America/Los_Angeles
Current time (UTC, cached as of): 2026-11-10T07:31:00.000Z
Query date (America/Los_Angeles): 2026-11-09

HTML: https://didcodexreset.com/news/1c051f7cb5324096bd7a413f.html

Content language: en
Localization state: sameLanguage

Source: Qwen · Published 9/20/2026, 07:46:21

Qwen announced Qwen-Image-2.1 as a unified image generation and editing model in the Qwen-Image series and said its weights are now open. The post describes a 7B architecture and links a blog, GitHub, Model Scope, and Hugging Face: https://t.co/tVntKOi7jy https://t.co/cRj66wCrWr https://t.co/64d7Ix6YFR https://t.co/njHSBXUbVS

Qwen said the model natively generates and edits RGBA layers for compositing and text editing in transparent images, supports up to 10 reference images, and provides precise local control while aiming to preserve fidelity for portraits and products. Follow-up posts from the same account said it can apply multiple region-specified edits in one pass and claimed improved text and portrait quality versus its predecessor.

Tags: Qwen, Qwen-Image, Image Generation, Image Editing, Open Weights

[View original post](https://x.com/Alibaba_Qwen/status/2101659302792679789)