Image releases of Oct 6-10, 2026: which have a Sume model id

Of the image releases dated October 6 to 10, 2026 in the pages I read, only Nano Banana 2.1 has its own model id on Sume: google/nano-banana-2.1. Qwen-Image-2.1-Turbo has no Sume id, and neither does Kroma v0.3.1, the Krea 2 turbo checkpoint. Sume's Image API lists 19 ids, and the only Qwen ones are qwen/qwen-image and qwen/qwen-image-max.

The cheapest way to stay current is to ask the catalog instead of a news page. GET /v1/images/models returns the ids your key can call today.

What shipped this week, and where does Sume stand?

Google's Gemini API changelog dates gemini-nano-banana-2.1 to October 6, 2026 and adds the wide ratios 1:4, 4:1, 1:8 and 8:1 at 1K, 2K and 4K. The same entry calls the older model deprecated with no shutdown date announced. A Kombitz roundup dates Qwen-Image-2.1-Turbo and Kroma v0.3.1 to October 10.

Does the old Nano Banana 2 id still work?

Yes. Sume's docs say google/nano-banana-2 and nano-banana-2 still work and run as Nano Banana 2.1, and the job stores the 2.1 id. The retired id is not in the catalog list, so a check against /v1/images/models will not find it even though a request using it succeeds.

How do I check a model id in code?

Fetch the list once and test membership. The third id below is a guess, so the script shows what a miss looks like.

import os
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()
ids = {m["id"] for m in r.json()["data"]}

for want in (
    "google/nano-banana-2.1",
    "qwen/qwen-image",
    "qwen/qwen-image-2.1-turbo",
):
    print(want, "listed" if want in ids else "not listed")

What if the model I want is not listed?

Pick the listed row that does the job you need rather than waiting. For a transparent PNG, ChatGPT Image 2.5 accepts background: transparent on Sume, and the Image API page says no other row takes that field. For a Qwen-style edit, qwen/qwen-image accepts up to 10 input references. Sume's docs do not say which Qwen version the qwen/qwen-image row runs, so do not assume it has the Turbo checkpoint's RGBA output.

Keep the request shape the same and swap only model; unsupported fields come back as 400 unsupported_parameter rather than being dropped, so a failed swap is loud.