Gemini 2.5 Flash Image shut down Oct 2: what to call on Sume instead
Google's gemini-2.5-flash-image, the original Nano Banana, was set to shut down Oct 2, 2026. Which Sume image ids to test as a replacement, and what to check.

Google's pricing page lists gemini-2.5-flash-image, the original Nano Banana, as deprecated with a shutdown date of October 2, 2026. If you called it directly, requests should now fail. On Sume you do not call that id: send a catalog id such as google/nano-banana-pro, or let sume/auto choose, and read the live list from GET /v1/images/models. This post says what to compare when you swap.
What Google says
The Gemini API pricing page, read 2026-10-04, says Gemini 2.5 Flash Image is deprecated and will shut down on October 2, 2026, and recommends moving to the newer Gemini 3.1 Flash Image models. It lists about $0.039 per output image. For comparison it lists Nano Banana Pro (gemini-3-pro-image) at $0.134 per image for 1K and 2K and $0.24 at 4K.
| Model | Output size | List price per image | Status |
|---|---|---|---|
| gemini-2.5-flash-image | up to 1024x1024 | $0.039 | Deprecated, shutdown Oct 2, 2026 |
| gemini-3-pro-image (Nano Banana Pro) | 1K or 2K | $0.134 | Listed |
| gemini-3-pro-image (Nano Banana Pro) | 4K | $0.24 | Listed |
What to send on Sume
Sume's catalog addresses models as org/slug, and the docs say the bare Image Router ids such as nano-banana-2 are accepted as aliases for their org/slug equivalents. Fetch the list and pick from it rather than guessing an id.
Read the descriptors on the Image API page: a request that sets a parameter the model does not list is rejected with 400 unsupported_parameter, not silently dropped. That matters on a swap, because aspect ratios and reference counts differ per family.
- Cheapest-first test: run your 20 most common prompts on the candidate ids at
n: 1. - Edits: set
aspect_ratio: "auto"so the reference's shape is kept. - Pin the id in config; do not rely on a floating alias that can change under you.
- Let
sume/autochoose only if you do not care which family runs; Sume does not disclose it.
Migration checklist
Replace the model string, then run your own prompt set. Compare cost from the usage.cost field on each response instead of the Google table, because Sume's billed amount is its own. Compare the failure mode too: a failed generation is not billed on Sume.
Check three behaviours your product depends on: text legibility, whether a person in the reference stays recognisable, and the output shape on edits. Those are the things that change between model families, and a list price will not show them.
curl "https://api.sume.com/v1/images/models" \
-H "Authorization: Bearer $SUME_API_KEY"
Not covered
Whether your old direct Google calls had other dependencies, such as interleaved text and image output, is outside what the Sume image route returns: it returns image URLs with usage data. See the Image API response format.
Sources
Related posts
More in Models
- Gemini 3.8 Flash TTS tops Hume's VoiceEQ board: run your own test
Hume's blog lists Gemini 3.8 Flash TTS atop its Real-World VoiceEQ board. Why a vendor-run board is only a lead, and how to run a blind A/B on your script.
- Omni Flash 1.1: GA in the Gemini API, Preview on Agent Platform
Gemini API release notes record gemini-omni-1.1-flash as GA on Aug 27, 2026; an Agent Platform page title still says Preview. What to check first.
- Omni's 1M-token context: how many ten-second edit turns fit?
Omni lists a 1,048,576-token context and 5,792 tokens per video second. If prior clips stay in context, 18 ten-second turns fit; Sume edits are stateless.
- Gemini Omni Flash GA: extension and interpolation vs Sume's inputs
Gemini Omni Flash is generally available with extension and interpolation between images. What Sume's gemini-omni-flash-1.1 catalog entry documents instead.
Written by Sume