Qwen Image 2.1: when you must show Built with Qwen
The Qwen Research License requires Built with Qwen or Improved using Qwen when outputs train a distributed model, and bars Qwen as a derivative's main name.
If you use Qwen-Image-2.1 or its outputs to create, train, fine-tune or improve an AI model that is distributed or made available, section 4(b) of the Qwen Research License says to prominently display "Built with Qwen" or "Improved using Qwen" in the related product documentation. Section 4(c) adds that "Qwen" cannot be the primary name of a derivative.
Clauses are quoted from the license file, read 2026-10-01. This summarizes the text and is not legal advice.
Which uses trigger the notice?
The trigger is creating, training, fine-tuning or improving an AI model that is distributed or made available. The clause covers the Materials and "any outputs or results therefrom", so training on generated images is within its wording. A model you keep private is outside the "distributed or made available" condition as written.
| Clause | Rule |
|---|---|
| 4(a) | Comply with export controls and applicable laws |
| 4(b) | Show "Built with Qwen" or "Improved using Qwen" in product documentation |
| 4(c) | Do not use "Qwen" as the primary name of a derivative; descriptive use such as "fine-tuned from Qwen Image" is permitted |
What naming is allowed?
The license permits reasonable descriptive use, with the example "fine-tuned from Qwen Image". Section 5(b) grants no trademark license except as needed for notice requirements or customary description of the Materials. A model called "Qwen Something" fails 4(c); "Something, fine-tuned from Qwen Image" is the pattern the text gives.
Does this apply to images I generate for ads or posts?
The 4(b) trigger is about building a model, not publishing pictures. Whether non-commercial scope in section 2 covers a given publication is a separate question; see the commercial-use post. Ask the licensor or a lawyer for your case.
Where does Sume fit?
Sume lists qwen/qwen-image and qwen/qwen-image-max as image ids, and the image docs say a parameter a model does not list is rejected with 400 unsupported_parameter. Nothing in these clauses is a Sume feature, and the cited Sume docs do not describe notice handling. See Qwen Image 2.1 API for what is hosted.
Sources
Related posts
More in Developers
- QwenImage21Pipeline in diffusers vs an HTTP image request on Sume
QwenImage21Pipeline runs Qwen-Image-2.1 locally on a GPU. Sume has no 2.1 id: a hosted call is one POST /v1/images with a listed model. Side-by-side.
- Qwen Image 2.1 LoRA redistribution: notice file and Hangzhou courts
Sharing Qwen-Image-2.1 weights or a LoRA needs the license copy, modified-file notices and a Notice file; disputes go to the Hangzhou courts. Clause summary.
- remove.bg API rate limit: 500 per minute, weighted by megapixels
remove.bg allows 500 images per minute at about 1 MP, less for larger inputs. Sume limits requests per minute per key and sends retry-after on 429.
- remove.bg bg_color replacement: Sume RMBG gives a transparent PNG
remove.bg's API has bg_color and bg_image_url. Sume RMBG 1.0 takes only an image_url and returns a PNG with alpha, so the new background is a second step.
Written by Sume