Skip to content
See the World Through ScienceA project of ALLATRA

Alibaba Released a Free Image Model, but Its License Bars Commercial Use

AI & Technology

Republish this story

Our work is licensed under Creative Commons BY-NC 4.0. You may republish this piece for free — with credit to ALLATRA Media and a link to the original, unedited beyond length trims, and not for commercial use.

Read the full license

The Alibaba Group headquarters building in Hangzhou, China, a glass office block with a white lattice facade and a canopy over the entrance plaza.
Alibaba Group's headquarters in Hangzhou, China. The company's Qwen team published the Qwen-Image-2.1 weights on Sept. 20 (illustrative)."Alibaba group Headquarters" by Thomas LOMBARD, designed by HASSELL (architects)[1], via wikimedia, CC-BY-SA-3.0 · CC-BY-SA-3.0

Alibaba's Qwen team published the weights of a new image model, Qwen-Image-2.1, on Sept. 20. The team posted the files on Hugging Face and ModelScope under a license that allows research and evaluation only.

The team's own repository describes the release as one model that does two jobs: making pictures from a written prompt and editing pictures it is handed. The same model can produce images with a transparent background, edit transparent layers, and lift a subject out of a photograph. Alibaba published no benchmark comparison with the weights, and no independent evaluation of the model has appeared.

The repository calls the release open-source. The document the weights ship under is not an open-source license. The Qwen Research License Agreement, dated Sept. 20, 2026, grants the right to use, copy, modify, and redistribute the model for non-commercial purposes only, and it defines non-commercial as research or evaluation. Commercial use requires a separate license from the company. Anyone redistributing the model has to pass the agreement on, mark any modified files, and label anything trained or fine-tuned on it as built with Qwen.

The same repository puts the generating half of the model at 7 billion parameters. It takes up to 10 reference images for a single composition, accepts local edits marked with circles, painted annotations or separate masks, and produces images at 2K resolution.

It also lists same-day support from five outside software projects that run the model, among them Hugging Face's Diffusers library and ComfyUI.

Sources

Spot an error?

Spot an error?

Report an error

Spotted a mistake on this page? Tell us what's wrong and our editors will take a look.

What kind of problem?

Only if you'd like us to be able to follow up. We won't use it for anything else.

We correct mistakes openly. Select any text to flag it. Fixes are logged under our Corrections Policy.

Report an error

Reporting on

Alibaba Released a Free Image Model, but Its License Bars Commercial Use

What kind of problem?

Only if you'd like us to be able to follow up. We won't use it for anything else.

We read every report. Corrections are logged publicly.