# Firebase's Imagen Deadline Moved Twice, Its Replacement Has One Too

> Source: <https://firerun.io/firebase-imagen-shutdown-delayed-2026/>
> Published: 2026-08-17 00:00:00+00:00

# Firebase's Imagen Deadline Moved Twice, Its Replacement Has One Too

Firebase's Imagen shutdown, pushed back once already, hits today. The replacement Google pointed everyone to already has its own October expiration date.

Firebase AI Logic’s Imagen models shut down today, August 17, 2026, according to Firebase’s own migration guide, which now describes the date as extended from a previous June 30, 2026 cutoff ([Firebase AI Logic, Imagen migration guide](https://firebase.google.com/docs/ai-logic/imagen-models-migration)). Any app still calling `ImagenModel`

on the Gemini Developer API or the Agent Platform Gemini API path stops generating images as of today.

This is the second time Google has moved this deadline. Firerun reported an Imagen shutdown on June 24 ([Firebase Drops Imagen: Migrate to Gemini Images Now](/firebase-imagen-shutdown-2026/), June 25, 2026), based on Firebase’s guidance at the time. That date didn’t hold. Google pushed it to June 30, then again to today. **Correction:** the June 25 article’s June 24 cutoff was premature; Firebase’s live docs now show the actual shutdown landing today. See the correction note added to that piece.

## What actually changed

- All Imagen models are deprecated and shut down as early as today, August 17, 2026, across both the Gemini Developer API and the Agent Platform Gemini API (formerly Vertex AI) (
[Firebase AI Logic, supported models](https://firebase.google.com/docs/ai-logic/models)). - The replacement path is unchanged from June: Gemini image models, internally code-named “Nano Banana,” reached through the same Firebase AI Logic client SDKs.
- The specific image-generating Nano Banana model has its own shutdown date already published: October 2, 2026 (
[Firebase AI Logic, FAQ and troubleshooting](https://firebase.google.com/docs/ai-logic/faq-and-troubleshooting)). Other Nano Banana variants and stable Gemini Live API 2.5 models aren’t affected by that date.

## Why the deadline moved

Firebase hasn’t published a public explanation for the extension. What’s visible is the pattern: an announced cutoff, developers migrating against it, then a quiet pushback in the docs rather than a blog post. The same page that names today’s date is the one flagging it as already extended once. There’s no separate changelog entry marking the change.

That matters for how teams should read Google’s Gemini deprecation dates generally. A date on a migration guide is Google’s current intent, not a fixed commitment. Firebase recommends using Remote Config or a server-side prompt template to swap model IDs without shipping a new app build. That’s advice for surviving a schedule that has already moved twice on this one feature.

## The replacement isn’t a resting place

Teams migrating off Imagen today land on a model that Google has already scheduled a shutdown for. `gemini-3.1-flash-image`

and `gemini-3-pro-image`

are the current Nano Banana tier for image generation, but the FAQ page pins the underlying image-generating Nano Banana model to an October 2, 2026 cutoff, six weeks after today’s Imagen deadline.

The practical takeaway is narrower than “migrate now and move on.” Point the model ID at a config value, not a hardcoded string, and check the supported-models page again in September. Hardcoding whatever Nano Banana variant is current today just sets up the same scramble in October that today’s Imagen deadline caused for anyone who hardcoded `ImagenModel`

in the spring.

## If you’re migrating today

The mechanics haven’t changed since June: swap `ImagenModel`

for a `GenerativeModel`

instance, replace `ImagenGenerationConfig`

with a standard `GenerationConfig`

carrying a response modality of `IMAGE`

, and keep the existing API key and project wiring. Firebase’s migration guide covers Swift, Kotlin, Java, Web, Flutter and Unity with before/after snippets.

Check the [supported models page](https://firebase.google.com/docs/ai-logic/models) before you ship, not a cached blog post, this one included.
