Text edits on Sume: use sume/auto or pin ideogram/ideogram-v4.5?

Pin ideogram/ideogram-v4.5 when your edit depends on how that model handles text, and use sume/auto when you only need a good image and do not care which family made it. Sume's Image API docs say Auto "selects the family for you and never discloses which one ran", that GET /v1/images/models does not list it, and that job.model stays sume/auto.

A text edit is a case where the family matters. Ideogram's launch post (read 2026-10-05) pitches 4.5 on precise edits and multi-turn editing. If you pick the model by that claim, you want that model on every call.

What changes between Auto and a pin

How to pin safely

Pinning is a one-line change: set model to the catalog id. The safe way to do it is to read the catalog first. GET /v1/images/models lists the models with their supported_parameters, and the per-endpoint call returns the definitive parameters and pricing. Store the id in one place in your code, not in every call.

The docs also say Sume accepts bare legacy ids like gpt-image-2 and nano-banana-2 as aliases for their org/slug forms. For a model you plan to keep, use the full id so your code reads the same as the catalog.

Pins and the catalog change

A pin makes you responsible for catalog changes. If a model id is retired or renamed, a pinned call fails where an Auto call would have adapted. Two habits cover this: test the id with a cheap request on a schedule, and keep a fallback id in your configuration so a change is a config edit, not a code deploy. Auto remains the fallback of last resort when you only need an image.

A rule of thumb

Use a pin when:

• The edit must keep letters exact (signs, labels, covers).
• You compare results across runs and need one model to hold constant.
• You budget from the per-image pricing line of one endpoint.
• You need the reference rule above.

Where Auto is fine

Auto suits early exploration, mood images, and cases where a wrong word in the output costs nothing. It also suits legacy callers: the docs say Image 1.0 is a compatibility alias for this Auto pipe.

Either way, read the catalog first. GET /v1/images/models/ideogram/ideogram-v4.5/endpoints returns the supported parameters and the pricing lines for the pin. A request that sets a parameter the model does not list returns 400 unsupported_parameter; Sume does not silently drop it.

The decision is not permanent. Many teams start on Auto while they learn what they need, then pin once one family clearly wins on their images. Because the docs say Auto never discloses which family ran, you cannot learn that from the responses; you learn it by running the same prompt with each pinned id and comparing. Do that comparison once, keep the winner, and note the date. Keep the scope of this advice in view. It rests on the Sume docs and the vendor pages named in the sources, read on 2026-10-05, and on nothing measured by Sume. Where a behavior depends on your own images, such as how a model redraws a certain typeface, run a small pilot at the low quality tier and judge the result yourself before you plan a batch. Write down the prompt, the model id and the quality tier you used, so the run can be repeated. When the catalog or the docs change, re-read them; the live catalog is the contract, and a post is only a snapshot of it.