Detect new AI image models: diff Sume GET /v1/images/models

The most reliable way to know whether an image model you read about is available on Sume is to ask the API: GET /v1/images/models lists every model id with its capabilities and the per-endpoint pricing is one call away. Run it on a schedule, save the ids, and diff against yesterday. New launches show up as added ids; retired ones show up as removed, so a pinned id that disappears does not surprise you in production.

This matters because launches are frequent and a vendor announcement is not a Sume release. Image models launch often, so check each new one against Sume's list instead of assuming it is there.

How fast is the field moving?

Dates below come from each vendor's own page on the day read.

What is the diff script?

It stores the id list in a local file and prints what changed since the last run. Schedule it with cron or any job runner:

import json
import os
import pathlib
import requests

r = requests.get(
    "https://api.sume.com/v1/images/models",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    timeout=30,
)
r.raise_for_status()
now = {m["id"] for m in r.json()["data"]}
store = pathlib.Path("image-models.json")
old = set(json.loads(store.read_text())) if store.exists() else set()
print("added:", sorted(now - old))
print("removed:", sorted(old - now))
store.write_text(json.dumps(sorted(now)))

What does the list not tell me?

It lists ids, not quality. Pair a new id with the comparison in run the same prompt through several models before you switch. sume/auto is deliberately absent from the list, and Sume never says which family it chose, so a diff cannot see routing changes under Auto. Legacy bare ids such as gpt-image-2 are accepted as aliases for their org/slug forms, so keep config on the full id.

Limits

Run the diff against the environment you deploy to. Capability descriptors can change without the id changing, so diff the supported_parameters too if you rely on one, such as the 16-reference ceiling on GPT Image 2.5.