Sume image catalog by price: 19 models from $0.005 to $0.264

The list in one sentence

Sume's image catalog lists 19 model rows on 2026-10-04, priced from $0.005 per image for Higgsfield Soul to $0.26375 for ChatGPT Image 2, and 14 of them accept reference images for edits. The sorted table below is the catalog's own pricing line, which already includes Sume's margin.

Two caveats come first. Nano Banana and Ideogram 4.5 rows are priced at their default tier, and ChatGPT Image 2.5 shows its high-quality 1024 display rate, so the real charge moves with resolution, quality and size.

What moves the price

• Nano Banana 2 costs $0.10 at 512 and 1K, $0.15 at 2K and $0.20 at 4K. Nano Banana Pro costs $0.1875 up to 2K and $0.375 at 4K.
• Ideogram 4.5 is $0.0375 at low, $0.075 at medium and $0.275 at high quality.
• ChatGPT Image 2.5 is token-billed, so quality, size and reference images all change the total.
• Imagen 4 Ultra costs the same $0.075 at 1K and 2K.

Using the list

Read the live numbers rather than this page when money is on the line: GET /v1/images/models lists every row and GET /v1/images/models/{id}/endpoints returns the billable line. The catalog also reports reference limits: 16 for ChatGPT Image 2.5, 5 for Ideogram 4.5, 10 for the other edit-capable rows, and 0 for the five text-only ones (Soul, Imagen 4 Fast, Imagen 4 Ultra, Recraft V4 and Qwen Image Max).

If you do not want to choose, send model: "sume/auto" and Sume's image router picks the family. Details are in the Sume Image API docs.

Three practical cuts

Soul at half a cent is text-only and returns one or four images per call, so it suits mood boards and idea sheets, not edits.

• Under three cents and editable: Grok Imagine and Qwen Image, both $0.025. Grok caps n at 1, Qwen at 4.
• Best mix of price and reference slots: Flux 2 Pro at $0.0375 takes up to 10 references, and Seedream 5.0 Lite at $0.04375 does the same.
• Text on the image: Ideogram 4.5 at its $0.075 medium tier takes up to 5 references; Ideogram V3 takes 10 at the same price but is the older model.
• Largest reference count: ChatGPT Image 2.5 accepts 16 references plus an optional mask_url, billed by tokens instead of a flat rate.

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.