Midjourney V8.2 API: does Sume list a Midjourney model?

No. As of 2026-09-29, Sume's image catalog does not list a Midjourney model of any version, so calling POST /v1/images with a Midjourney id returns 404 model_not_found. This post makes no claim about Midjourney or its versions; it covers only what Sume lists and how to check the list yourself.

How do I see exactly which image models Sume lists?

Call the catalog. It is the live source of truth: the list only includes models whose provider is configured at that moment, so a copy of it in a blog post goes stale.

curl "https://api.sume.com/v1/images/models" \
  -H "Authorization: Bearer $SUME_API_KEY"

What does a Midjourney id return?

An id that is not in the catalog fails before anything is generated. The response is 404 with code model_not_found, the message Unknown image model id., and a catalog_url pointing at /v1/images/models. Nothing is billed.

curl -X POST "https://api.sume.com/v1/images" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "midjourney/v8.2", "prompt": "a lighthouse at dusk"}'

Which model families does the catalog list instead?

Ids use an org/slug shape. These are examples from the catalog code, not the full list; run the call above for the current one.

Can Sume pick a model for me?

Yes: send model: "sume/auto" and Image Router chooses. The response echoes sume/auto, not the family that served it. The behavior is described in Image API.

curl -X POST "https://api.sume.com/v1/images" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "sume/auto", "prompt": "a lighthouse at dusk"}'

What should I compare before switching to a listed model?

Read each candidate's record instead of guessing. GET /v1/images/models/{model_id}/endpoints returns the definitive parameter set and the pricing lines, and Sume's margin is already in cost_usd, so cost_usd times n is the price. Parameters differ by model: some take reference images, some are text-only, and each lists its own aspect ratios. A request that sets a parameter the model does not list returns 400 unsupported_parameter.

Billing is all-or-nothing. A completed generation is billed in full, a failed or cancelled one is not, and a request that fails validation never starts. That makes a trial call across two or three listed models a cheap way to see which output fits your use.

Which parameters does every catalog row publish?

Each row's supported_parameters uses typed descriptors, so you can compare models without trial calls.

What if a model I used disappears from the list?

The list only includes models whose provider is configured at that moment. If a model you used yesterday is missing from today's response, treat that as the catalog changing, not as an error in your request, and pick from what the call returns.