FLUX 3 Image edit API: no mode field, just add references

FLUX 3 Image has one endpoint, POST /v1/flux-3-image, and no mode field: add up to ten reference images and the request edits, restyles or combines them; leave them out and it generates from the prompt. Sume also has one image route, POST /v1/images, but whether references are accepted depends on the model's catalog entry, and text-only models reject them.

BFL facts are from its FLUX 3 Image reference; Sume facts from the Image API and catalog code, read 2026-10-01.

How does FLUX 3 Image decide to edit?

By the presence of images: one URL or base64 string, or a list of 1 to 10. Each image must be at least 256 x 256 and at most 16 megapixels, and each base64 payload at most 20 MB. You describe in prompt how to use them. Sending a mode field is an unknown field, which the spec says returns 422.

How does Sume decide?

On Sume you put reference images in input_references, and the model has to allow them. In the catalog code, edit-capable entries are flagged image_edit: true, and text-only entries such as the Imagen 4 ones carry the constraint "text-to-image only; image_urls not accepted". Models whose input_references descriptor is {"min": 0, "max": 0} reject references.

What does a Sume edit request look like?

Pick an edit-capable model from GET /v1/images/models, put the picture in input_references as a public HTTPS image_url, and describe the change in prompt. Set aspect_ratio to auto so the output matches the reference; the docs say omitting the field is not the same as auto. Note that on Sume mode means something else: sync, async, subscribe or webhook control how you wait for the result.

Is FLUX 3 Image available on Sume?

Not in the catalog code I read: the FLUX entries are flux-2-pro and flux-2-flex, both edit-capable. For edits with those, see FLUX.2 multi-reference editing, and for the wider comparison what Sume serves for FLUX 3.