Premium image models on Sume ranked: 7 to 27 cents a picture

Ranked by Sume's billable catalog price, the higher-priced image models run from 7 cents (ChatGPT Image 2.5 at its high rate, and FLUX.2 Flex) to 27 cents (ChatGPT Image 2). Nano Banana 2.1 and Qwen Image Max are both 10 cents, and Nano Banana Pro is 19 cents. The ranking is by price only; the Sume catalog gives no quality ordering, so use it to shortlist, not to decide.

The ranked list

Sume bills the catalog list per image times 1.25, rounded up to the cent. The catalog's list for some models is a display rate, such as high quality at 1024 pixels, so these are the numbers to compare at that size, not a guaranteed price for every call.

Three things the ranking hides

• Two of these rows are text-to-image only, with no reference images: Imagen 4 Ultra and Qwen Image Max. If your job is an edit, strike them.
• Resolution can raise the cost. Nano Banana 2.1 and Pro scale at 0.5K, 2K and 4K, and Pro is flagged as higher at 4K.
• GPT Image 2.5 changes with quality: the 7-cent figure is high, low is about 1 cent, and auto reserves max.

How to shortlist

Use the table to cut the field to two or three models for your task, then compare them on one prompt. For a hero image, a premium row such as ChatGPT Image 2.5 high at 7 cents or Nano Banana 2.1 at 10 cents is a sensible start. For a 4K output, read the Nano Banana notes before you budget, because the admission price scales. See the model list with GET /v1/images/models on the Image API page, and pin the model id instead of using sume/auto when you are comparing.

Check it before you run it

Every figure above is a catalog list price times 1.25, rounded up to the cent, as of 2026-10-08. Catalogs change, so before a large batch, read the current model entry in the docs and recompute the one line that matters for your case. Write the arithmetic next to the job in your own notes: list rate, seconds or characters, multiplier, rounding. If the result differs from the wallet charge by more than a cent, the catalog entry has changed, and the docs page is the place to find out why.

Run one small job first. Submit a single request with the pinned model id and the settings in the tables, poll the returned polling_url until it finishes, and compare the charge with your estimate. Then scale up. Using a pinned id for the test matters, because sume/auto never names the family, so you cannot tie its charge to the row you priced.