Qwen-Image-2.1 is non-commercial: hosted options
Qwen-Image-2.1 has open weights but a Qwen Research License that is non-commercial unless licensed. Check the license first, then the hosted catalog.

Qwen-Image-2.1 ships open weights, but according to OrcaRouter it falls under the Qwen Research License, which is non-commercial unless you obtain a license. If you plan to sell what you generate, resolve the license before choosing between self-hosting and a hosted API.
What the comparison post reports
OrcaRouter's comparison lists the following for a September 20, 2026 release. This is a third-party summary; read the license text yourself before relying on it.
| Item | Value |
|---|---|
| Release | Sep 20, 2026 |
| Weights | Open |
| Size | 7B |
| Capabilities | Text-to-image and editing |
| License | Qwen Research License, non-commercial unless licensed |
| Text-to-image ranking | Elo 1,034, rank 18 of 166 |
Open weights are not open use
The weights being downloadable does not grant commercial rights. A research license usually restricts selling outputs or serving the model to customers. If your use is a paid product, an ad or a client deliverable, get written confirmation of what the license allows.
Hosted route
Sume's Image API lists models through GET /v1/images/models, with per-model supported_parameters and, on the per-endpoint records, pricing lines. Sume's docs do not name a Qwen id, so use the catalog to see what is currently available rather than assuming. Pricing lines are what your wallet is charged, with Sume's margin already included.
Hosted models still carry their own terms. Read the model page and the Sume terms for commercial use before you ship.
Decision checklist
- Is the use commercial? If yes, confirm license terms in writing.
- Do you need transparent output? Check whether the catalog model supports
background: transparent. - Do you need edits? Check that
input_referencesallows at least one image. - Compare cost per image after the license question is settled.
Sources
Related posts
More in Models
- Repeated image edits without drift: a native-resolution test plan
Recraft says Ideogram 4.5 is built for stable repeated edits at native resolution without downsampling. A five-pass test to check any edit model for drift.
- Scribe v2, Realtime or Medical: which model for interview clips
ElevenLabs lists Scribe v2 with up to 32 speakers and 90+ languages, Realtime at about 150 ms, and Medical with 35% fewer errors. Which fits interview clips.
- Seedance 2.5 on BytePlus excludes the US: check access first
BytePlus lists Seedance 2.5 on ModelArk in supported markets excluding the United States. What that means for a US team, and how to read the Sume catalog.
- Self-host MiniMax H3 or use an API: a decision table
MiniMax released H3 as an open-weight omni-modal model on Jul 31, 2026. A decision table for running it yourself versus calling a hosted video API.
Written by Sume