Which Sume image model after this week's launches: lookup by need

After this week's image launches, the short answer on Sume is: Imagen 4 Fast for cheap text-only drafts, Nano Banana 2.1 for most photo edits, ChatGPT Image 2.5 when you need more than 10 references or a transparent background, and nothing for Hy Image 3.5 Preview, which Sume does not list. Prices are Sume list as of 2026-10-08.

Lookup by need

Each row uses limits from the Sume catalog. Reference counts come from input_references, and the 4K row from the resolution descriptor.

What the vendors say

Google's pricing page describes Nano Banana 2.1 as an update to Nano Banana 2 with improved visual quality, character consistency and text rendering at 1K, 2K and 4K. OpenRouter's page describes Tencent Hy Image 3.5 Preview as a unified generation and editing model taking up to 20 reference images, with output up to 4K. Google's Imagen page says Imagen models are shut down in the Gemini API. These are vendor statements; test on your own images.

Check the list on the day

The catalog changes. Run GET /v1/images/models before you pin an id. A retired id such as nano-banana-2 runs as nano-banana-2.1 on Sume, so old code keeps working, but you should update the id so your logs show what actually ran.

Reading the table

The table is a starting point, not a verdict. Every row says what the model can do in Sume's catalog, not how good its pictures are for your subject. Run 10 of your own prompts on the top two rows for the need, which costs under $2 on most of them.

Hy Image 3.5 Preview sits on the table only to say it is not there. If it appears in the catalog later, it will show up in GET /v1/images/models. Until then, this post on what to call instead lists the closest rows.

How this week's launches fit

Google's pricing page lists Nano Banana 2.1 and calls it an update to Nano Banana 2. Sume lists the same family as google/nano-banana-2.1 and runs the old id as the new one. Tencent's Hy Image 3.5 Preview appeared on OpenRouter with a $1.60 per million token price, and Sume's catalog does not carry it. Imagen 4 is shut down in the Gemini API according to Google, yet two Imagen rows exist in Sume's catalog today.

These are three different situations: an update that Sume follows, a new model that Sume does not list, and a retired Google model that Sume still lists. Reading the catalog each day is the only way to know which applies to the id you plan to use.