Qwen-Image-2.1
Image modelOpen weightsCard summarizedA unified text-to-image generation and image editing model in the Qwen family
Source: Qwen-Image-2.1 model card (Hugging Face) and GitHub README
Where to call it
- Alibaba Cloud Model Studio
qwen-image-2.1-pro - Self-host (weights)
Qwen/Qwen-Image-2.1
Good for
- Transparent assets: stickers, product cut-outs, layered designs
- Multi-reference editing that preserves people and products
- Typography and portrait photography
Overview
Alibaba · 01Qwen-Image, front
Qwen-Image 2.1
A unified text-to-image generation and image editing model in the Qwen family
Typical price
$0.04 / image
qwen-image-2.1-pro (API)
- Max resolution
- 2K native (e.g. 2048×2048, 2752×1536)
- Aspect ratios
- 7 options
- Modes
- text-to-image, image editing
- Released
- 20 Sep 2026
Qwen-Image-2.1 is an open-weights model that unifies text-to-image generation and editing with a 7B visual generation component (32 single-stream DiT layers). It natively generates and edits transparent RGBA images, takes up to 10 reference images with local edit marks or masks, and improves typography, portrait lighting and detail. A Turbo checkpoint runs in 8 steps. model card
qwen-image-2.1-pro- Typical price
- $0.04 / image
- Max resolution
- 2K native (e.g. 2048×2048, 2752×1536)
- Aspect ratios
- 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16
- Weights
- Open · Qwen Research License (non-commercial)
- Released
- 20 Sep 2026
- text-to-image
- image editing
- multi-image reference (up to 10)
- transparent (RGBA) generation and layer editing
- subject extraction
- local edits via circles, painted marks or masks