Firebase Drops Imagen: Migrate to Gemini Images Now

Firebase AI Logic’s Imagen models went dark on June 24, 2026. No grace period extension, no soft cutoff. They’re gone. Any app still pointing at an ImagenModel instance stopped generating images yesterday. The replacement is Gemini’s “Nano Banana” image models, which have been GA for weeks, and Firebase published the migration guide before the shutdown.

Key Takeaways

• All Imagen models shut down June 24, 2026, across both the Gemini Developer API and Vertex AI paths (Firebase Release Notes, June 2026).
• The direct replacement is Gemini 3.x image models (codenamed Nano Banana), available through Firebase AI Logic (Firebase AI Logic docs, 2026).
• Migration is a small API swap: ImagenModel → GenerativeModel, with a response modality set to IMAGE (migration guide).
• Two model tiers cover most use cases: gemini-3.1-flash-image for speed, gemini-3-pro-image for quality.

The two replacement models

Google shipped two Nano Banana variants as GA before pulling Imagen:

gemini-3.1-flash-image (Nano Banana 2) is for speed and high-volume workloads. It also accepts video input, which Imagen never did. Use it for anything that needs throughput over maximum image fidelity.

gemini-3-pro-image (Nano Banana Pro) is the quality tier, built for professional asset production with stronger reasoning behind the generation. If you were using Imagen 3 for polished output, start here.

Both run through the same Firebase AI Logic client SDK you already have. No new dependency.

What the migration actually changes

The code change is smaller than the announcement suggests. You swap the model class and update the generation config:

• Replace ImagenModel with a GenerativeModel pointed at a Nano Banana model ID.
• Replace ImagenGenerationConfig with a standard GenerationConfig, adding a response modality of IMAGE.
• API key and project wiring stay the same.

Firebase’s migration guide covers every SDK (Swift, Kotlin, Java, Web, Flutter, Unity) with before/after snippets. The guide went live before the shutdown date, so there’s no excuse for being caught without it.

Why Imagen is gone

Google deprecated Imagen across the full platform, not just Firebase. The deprecation covered both the Gemini Developer API and Vertex AI Gemini API paths simultaneously. The “Nano Banana” codename is Google’s internal label for the Gemini 3.x image-generation capability, the same model family behind Imagen’s replacement in Google AI Studio and Vertex AI.

The Gemini image API is now the single path for AI-generated images across Google’s stack. There’s no parallel Imagen track to fall back to.

If your app is already broken

Check the Firebase AI Logic supported models page for the current model ID list. Don’t rely on a cached blog post. Model IDs shift with new releases, and hardcoding a version you haven’t verified recently is how you get broken again in six months.

The migration guide is the authoritative starting point. Most apps will be back up in an hour.