Imagen 4 Ultra on Sume: 5 aspect ratios, 1K or 2K, $0.075

Answer

On Sume, Imagen 4 Ultra (google/imagen-4-ultra) accepts five aspect ratios (1:1, 16:9, 9:16, 4:3, 3:4), two resolution tiers (1K and 2K), and n from 1 to 4, at $0.075 per image for both tiers as of 2026-10-04. It is text-to-image only: the input_references range is 0 to 0.

The price does not rise at 2K, so choose 2K whenever the image is going to be cropped or printed. Imagen 4 Fast is the cheaper sibling at $0.025, with five ratios and no resolution parameter.

What it cannot do

• No 4:5 or 3:2. Ask for either and the request is rejected with a 400; use 3:4 and crop, or pick a model that lists 4:5.
• No photo editing. A request with input_references is rejected because the model advertises none.
• No quality or background parameter. Both return 400 on this row.
• No seed. No catalog model advertises one in v1.

Request

A valid call names the model, one of the five ratios and a tier. Everything else is left out.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"google/imagen-4-ultra","prompt":"a ceramic teapot on a linen cloth, soft daylight","aspect_ratio":"4:3","resolution":"2K"}'

A 2K ultra image can take longer than the 30-second sync wait. If so, the response is a 202 job envelope, and the result arrives from the job result endpoint. See the Sume Image API docs for the full parameter table.

When to choose it

Pick Imagen 4 Ultra when you want a text-only render at 2K without paying more than the 1K price, and your layout is one of the five ratios. For a tall 4:5 social crop, the catalog rows for Seedream, Flux, Recraft, Nano Banana, GPT Image and Ideogram all list 4:5.

For a draft pass at a fifth of the money, run Imagen 4 Fast at $0.025 with the same prompt and ratio, then rerun the keepers on Ultra. Three drafts at Fast plus one Ultra final costs $0.075 + $0.075 = $0.15 per finished image.

Before a large run

Prices and descriptors change when the catalog changes, so confirm them before you spend. Call GET /v1/images/models/{id}/endpoints for the row you plan to use and read its pricing line and supported_parameters; both come back in one response.

Then run a pilot of three to five images and read usage.cost on each response. Multiply by your planned count for a forecast you can trust. Completed generations are billed in full and failed or cancelled ones are not, so a pilot that errors costs nothing.

For big batches, use mode: "async" or mode: "webhook" with a public HTTPS webhook_url, so no request waits on the 30-second sync limit. Poll GET /v1/jobs/{id}/status and fetch GET /v1/jobs/{id}/result when the job completes.