New image model out? Check Sume's catalog before you integrate

Check the catalog, not the launch post. Sume lists a model only when GET /v1/images/models returns it for your key; a model that shipped elsewhere may not be there yet. The Image API docs and OpenAPI schema I read on 2026-10-02 list FLUX.2 Pro and Flex and Ideogram V3, and do not list FLUX 3 or Ideogram 4.5.

Higgsfield's changelog shows FLUX 3 Image added 2026-10-01 (complex prompts, more accurate text, text-to-image and image-to-image up to native 4K) and Ideogram 4.5 added 2026-09-30 (reference edits and text replacement that preserve layout, at 1K or 2K) (Higgsfield changelog, read 2026-10-02). Those are Higgsfield facts, not Sume facts.

What does Sume list today?

The model enum on POST /v1/images in the OpenAPI schema includes these families, among others. Use the catalog call for the live answer.

What do I use while a new model is not listed?

Pick the closest listed model for the job. For text-heavy ads, test a higher quality setting on your own text, and note that ChatGPT Image 2.5 accepts up to 16 reference images and an optional mask_url. Or send model: "sume/auto" and let Image Router choose; Sume does not disclose which family ran.

How do I avoid hard-coding a missing id?

A wrong id fails the request, so keep the id in config and run the catalog read in CI. A 404 model_not_found means the id does not exist in your workspace, and the docs point you to /v1/catalog to verify ids. See Image API and Generation admission for the error table.