Midjourney V8.1 made HD the default: which Sume image models list 2K?

Midjourney V8.1 made HD the default, saying HD mode is "3x faster and 3x cheaper". If you want higher resolution from the Sume image API, resolution is a per-model setting, not a global switch. The resolution field takes a tier (512, 1K, 2K or 4K), and a model accepts only the values its catalog descriptor lists. A short script over GET /v1/images/models tells you which models advertise 2K or 4K today.

What the vendor says

From the V8.1 Alpha post.

What the Sume docs say

The Image API docs define resolution as a normalized tier (512, 1K, 2K, 4K). They add that a model accepts only the values its catalog descriptors list, so read supported_parameters before pinning a tier. Descriptors are typed: enum for an allowlist, range for integer bounds, boolean for supported or not.

A tier a model does not list is rejected with 400 unsupported_parameter, so you find out at the cost of a request, not an image.

The script

It lists every model whose resolution descriptor is an enum that includes the tier you want. Models with no resolution descriptor are skipped, because the catalog does not advertise the field for them.

import os, requests

WANT = "2K"
r = requests.get(
    "https://api.sume.com/v1/images/models",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    timeout=30,
)
for m in r.json()["data"]:
    res = m.get("supported_parameters", {}).get("resolution")
    if res and res.get("type") == "enum" and WANT in res.get("values", []):
        print(m["id"], res["values"])

Using the answer

Pin one of the listed models and send the tier with the request. Higher tiers are among the configurations most likely to run past the 30 second wait, so expect a 202 job envelope and read the image from the job result. See the Jobs and results docs.

Cost follows the endpoint pricing line, so check GET /v1/images/models/{id}/endpoints before you move a whole batch to a higher tier.

Notes

Two cautions:

• Midjourney's "HD" and Sume's "2K" are different things; compare outputs, not labels.
• The catalog changes. Re-run the script when you plan a release, or diff its output weekly.