Qwen Image 2.1 API: open weights, and what Sume lists

Qwen-Image-2.1 is an open-source model you run yourself, and as of 2026-09-29 its Hugging Face page says no Inference Provider deploys it. Sume's image catalog does not list a model named 2.1. The Qwen models Sume lists are qwen/qwen-image and qwen/qwen-image-max.

The model facts come from the Hugging Face card; the Sume side from the Image API docs and the catalog code.

What does the Qwen-Image-2.1 model card say?

The card describes a unified text-to-image and image-editing model with 7B parameters in its visual generation component. It lists native transparent (RGBA) generation, up to 10 reference images for editing, and the license qwen-research. Its example code maps seven aspect ratios to sizes, such as 1:1 at 2048×2048 and 16:9 at 2752×1536.

Those are the card's numbers for weights you host. They are not Sume's parameters; Sume takes the aspect ratios its own catalog lists for each Qwen model.

Is Qwen Image 2.1 available through Sume?

Not under that name. The catalog is the source of truth, and it is one call to read:

curl "https://api.sume.com/v1/images/models" \
  -H "Authorization: Bearer $SUME_API_KEY"

How do I call a Qwen model on Sume?

qwen/qwen-image is estimated at $0.03 per image on Sume and qwen/qwen-image-max at $0.10. Both take aspect_ratio from the catalog; qwen/qwen-image lists 1:1, 16:9, 9:16, 4:3, 3:4, 4:5, 5:4, 3:2, 2:3, 21:9, 9:21, 1:2, 2:1. Sume does not promise exact pixel sizes, so the model card's 2048×2048 figure does not carry over.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen-image",
    "prompt": "A hand-lettered chalkboard menu, blue and white",
    "aspect_ratio": "3:4"
  }'

Which license does the Qwen-Image-2.1 card list?

The card's license field reads qwen-research. That is the license for the open weights you download; this post makes no claim about terms for hosted calls. If you plan to ship the weights inside a product, read the license text itself before you build.

Can I get a transparent PNG from Sume?

Sume's docs do not list transparent output for any Qwen model, so do not rely on it. Composite your own background or cut the subject out after generation. If you need native RGBA today, the model card's route is running the weights yourself, under the qwen-research license terms.

Will Sume add Qwen-Image-2.1?

The docs make no promise either way, so this post makes none. If it is added, it appears in GET /v1/images/models with its parameters and its price.