gemini-omni-flash-preview shuts down 2026-09-30: which line to change, which id to use, how to verify
Google lists gemini-omni-flash-preview under Deprecated models on the Gemini API deprecations page: released 2026-06-30, earliest shutdown 2026-09-30, recommended replacement gemini-omni-1.1-flash. This page covers the migration only - which id expires, which id replaces it, which single field in your request changes, and how to check it afterwards. QCode does not currently offer the Gemini series; please use Claude, GPT or Chinese models.
Updated 2026-10-08
Hit your subscription limit or keep getting 429s? Switch to an API billed per token and keep working with one key.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four cells on Google's table
earliest shutdown date
The Shutdown date cell for this row reads 2026-09-30. The same page adds that the dates are the earliest possible retirement days and that the exact shutdown date is communicated in advance. Treat it as a deadline, not as a promise that nothing moves before that day.
the only id in Recommended replacement
That cell names one id, gemini-omni-1.1-flash. The same table gives it a release date of 2026-08-27 and no shutdown date. Google publishes no mapping of capability, quota, context or price between the two rows, so equivalence is yours to measure.
the migration surface in the request
The replacement cell swaps an id and says nothing about a new endpoint, new authentication or a different payload shape. If the model name is the only place you hard-code it, the edit is one value - our reading of the table, not a sentence Google wrote.
Gemini series on QCode
QCode does not currently offer the Gemini series; please use Claude, GPT or Chinese models (e.g. claude-sonnet-5-5, gpt-6.1-sol, glm-5.3).
What “shut down” means on that page
The page draws its own border: it lists known deprecation schedules for stable and preview models in the Gemini API - one calling surface, not "all of Gemini". It also separates two words. Deprecation is the announcement that support has ended and shutdown is coming; shutdown means the model is completely turned off and the endpoint is no longer available. The dates carry the same caveat: earliest possible, exact date communicated in advance. This page writes from the text of that table only - the gray row that flags an already-shut-down model is a visual device a text archive cannot read, so nothing here states whether the entry answers at this moment.
Which version of the table this page reads
Our archive of the Google page was captured 2026-09-21, and the page itself says Last updated 2026-09-17 UTC. At capture time the deprecated row already carried all three cells: released 2026-06-30, shutdown 2026-09-30, replacement gemini-omni-1.1-flash. Google adds that announcements go to the Release notes page and that this table only tracks the dates. Our check date is 2026-09-20, i.e. 10 days before 2026-09-30 - our own subtraction, not an official sentence. If the table changes, every number here needs re-verifying.
Timeline
2026-06-30: the table's Release date cell for gemini-omni-flash-preview. Release to the earliest shutdown date 2026-09-30 is about 3 months - our subtraction, not Google's wording.
2026-08-27: the replacement id gemini-omni-1.1-flash appears in the table with No shutdown date announced. It landed 34 days before the old id's earliest shutdown day (our arithmetic), so nobody has to wait for the successor.
2026-09-30: the earliest shutdown date in the table. QCode does not currently offer the Gemini series.
Confirmed vs caution
Officially confirmed
Confirmable from the archive we hold (captured 2026-09-21, page self-dated 2026-09-17): release 2026-06-30, earliest shutdown 2026-09-30, a replacement cell naming gemini-omni-1.1-flash and nothing else, that replacement released 2026-08-27 with no shutdown date, and the definition that a shut-down model has its endpoint unavailable. All from that table, none inferred.
⚠️ Caution
Four things this page does not write. 1) Whether the two entries match on capability, context length, quota or price - the table has no such column. 2) What the failure looks like on the day: only “no longer available”, no code, no timing. 3) Anything past this one id and this one calling surface; this is not “Gemini being wound down”. 4) Whether QCode will list gemini-omni-1.1-flash: our 2026-09-20 snapshot does not contain it, and “coming soon” is not a fact.
The id from the table vs our catalog as it stands
Official replacement: gemini-omni-1.1-flash
The only id Google names, released 2026-08-27 with no shutdown date announced. It is a separate entry on Google's endpoint, and nothing in the table says the two are equivalent in capability, limits or price.
QCode catalog state
QCode does not currently offer the Gemini series, so neither gemini-omni-flash-preview nor gemini-omni-1.1-flash is offered; please use Claude, GPT or Chinese models (e.g. claude-sonnet-5-5, gpt-6.1-sol, glm-5.3).
Four steps: search, swap, verify, keep evidence
First, search everywhere for the string gemini-omni-flash-preview: code, .env, gateway route tables, eval scripts, fallback lists - each leftover is one sudden failure. Second, change the value only: point model at gemini-omni-1.1-flash; the table mentions no change to endpoint, authentication or payload shape. Third, verify: send one minimal request on the new configuration, confirm it answers and that your usage shows the entry you intended, then run your longest production-shaped input. If you call through an aggregator, check that service's model list first (QCode does not currently offer the Gemini series). Fourth, keep the pre-change log and response, so a later difference points at the model or at your config.
On QCode
QCode does not currently offer the Gemini series — neither gemini-omni-flash-preview nor gemini-omni-1.1-flash; please use Claude, GPT or Chinese models (e.g. claude-sonnet-5-5, gpt-6.1-sol, glm-5.3). The shutdown date and replacement id on this page only restate Google's official deprecations table; models and prices are on /models and /pricing.
FAQ
Do we have to wait until 2026-09-30?
No, and don't. The table calls its dates the earliest possible retirement days and says the exact shutdown date comes with advance notice. Doing the edit this week is safer than leaving it to 2026-09-30.
How do I know the change actually took?
Three checks: the string is gone from code and config; a minimal request on the new setup answers normally; your usage records show the entry you intended. The first two can be done any day before 2026-09-30.
Will the new id produce the same output?
The table compares nothing, so this page does not write the word “same”. Run your own samples, especially long multi-turn context and anything on the tool-calling path.
What error comes back after the shutdown?
The documentation goes only as far as “the endpoint is no longer available”: no status code, no payload shape. Do not hard-code a branch around a guessed 4xx - keep logs from either side of the date and watch your own alerts.
We call this id through QCode. Do we lose it?
QCode does not currently offer the Gemini series, so this id is not offered on QCode; please use Claude, GPT or Chinese models (e.g. claude-sonnet-5-5, gpt-6.1-sol, glm-5.3). Google's table gives its earliest shutdown date as 2026-09-30.
Are other ids on that line expiring too?
This page looked at two rows in the Omni Flash section and nothing else, so do not read it as the whole Gemini line being retired. For any other id, find its own row on that table and see whether the Shutdown date cell holds a date or the words No shutdown date announced.
Sources
Google, “Gemini deprecations | Gemini API | Google AI for Developers”, https://ai.google.dev/gemini-api/docs/deprecations - archive captured 2026-09-21, the page itself says Last updated 2026-09-17 UTC; the two rows in the Gemini Omni Flash section are this page's only external evidence. Dates and ids follow the latest official table; nothing is extrapolated here.
Close that one line first, then choose a home
QCode does not currently offer the Gemini series; please use Claude, GPT or Chinese models. Models and prices are on /models and /pricing.
Related
Gemini 4 tracker: pre-training confirmed, no date given
What Google actually said about the next model, and what it did not.
Codex 401 Unauthorized: Causes and Fixes
codex login, switching to an API key, auth.json, env_key vs requires_openai_auth, and 401 checks for QCode.
ChatGPT Pro usage limits in Codex (October 2026)
Pro 200 keeps the old allowance through 2026-10-29; Pro 500 and Ultrafast at 8x / 6x.
Not affiliated with Google. Dates and ids are transcribed from a public page captured 2026-09-21 and self-dated 2026-09-17; “3 months”, “34 days” and “10 days” are our own subtractions of two dates in the table. QCode does not resell Google's endpoint and does not quote its prices.