How many reference images does each Sume image model take?

The input_references ceiling on the Sume image API is 16 on ChatGPT Image 2.5 (both ids), 5 on Ideogram 4.5, 10 on eleven other models, and 0 on five text-to-image-only models, read 2026-10-03. The number is published per model as the input_references range descriptor in GET /v1/images/models; the Image API page calls a {"min": 0, "max": 0} descriptor text-to-image only.

Reference URLs must be public HTTPS. Localhost, private-network and non-HTTPS URLs are rejected before submission.

Ceiling per model

Sorted from the highest ceiling down.

What counts as a reference

On most rows a reference is an image the model may edit or take style, identity or layout from. Ideogram 4.5 is different: with references the first image is the one edited and up to four more guide the result, which is why its ceiling is 5. A request with more references than the ceiling is an input error.

Pick by how many images the job needs

Choose by the descriptor, not by memory, because the catalog is the source of truth when a row changes.

• One photo to edit: any row with a ceiling of 1 or more (everything except Soul, Qwen Image Max, Imagen 4 Fast and Ultra, and Recraft V4).
• A product plus a logo plus a style frame (3 references): all ten-reference rows work.
• A character sheet or a catalog set of 11 to 16 pictures: ChatGPT Image 2.5, the only row above 10.
• Text only: Soul, Qwen Image Max, Imagen 4 Fast, Imagen 4 Ultra and Recraft V4 reject references.

Read it in code

List every model with its ceiling and route on that number. A call that is still running after the 30-second wait returns 202 with a job envelope instead of 200, so check the status code before reading data; see Jobs and results.

import os
import requests

headers = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
models = requests.get("https://api.sume.com/v1/images/models", headers=headers, timeout=30).json()["data"]
needs = 12
for m in models:
    top = m["supported_parameters"]["input_references"]["max"]
    if top >= needs:
        print(m["id"], top)