Midjourney edit model image references: 4 vs Sume's ranges

Midjourney's V8.2 edit model generates with other images, replacing omni-reference, with up to 4 image references at once. Sume has no single number: the reference ceiling is set per model, and GPT Image 2.5 accepts up to 16 references through input_references.

The Midjourney side is from its Aug 27, 2026 announcement; the Sume side from the Image API docs, both read 2026-10-01.

What does Midjourney's edit model change?

The announcement lists instruction edits, generating with other images (replacing omni-reference) with up to 4 image references, inpainting and outpainting, and use with personalization, moodboards and srefs. Access is through the attach-to-prompt bar, the lightbox edit button, the edit tab, or --edit url... on Discord. It does not describe an API.

How many references does Sume accept?

It depends on the model you select. The docs say the parameters a model accepts are published as capability descriptors, so you can read them before calling.

What happens if I send too many or unsupported inputs?

A request that sets a parameter the selected model does not list is rejected with 400 unsupported_parameter rather than silently dropped. Reference URLs must be public HTTPS. So pick the model by its descriptor, not by the count you remember from another tool.

const res = await fetch("https://api.sume.com/v1/images/models", {
  headers: { Authorization: "Bearer " + process.env.SUME_API_KEY },
});
const body = await res.json();
// Read each model's supported_parameters.input_references range.
console.log(JSON.stringify(body, null, 2));

Does Sume replace omni-reference?

Sume does not claim to reproduce Midjourney's behavior. What you get is reference-guided generation: put each reference in input_references as an image_url item and describe in the prompt what each one is for. For style references specifically, see Midjourney sref on Sume.