Imagen 3 & 4 Shut Down June 24: Migrate to Gemini Image (2026)

June 24, 2026. That is the shutdown date for every Imagen model on Firebase AI Logic — imagen-3.0-generate-002, imagen-4.0-generate-001, imagen-4.0-ultra-generate-001, imagen-4.0-fast-generate-001. All of them. If you have been putting off this migration, you have run out of runway.

The replacement is Google's Gemini Image models — internally called "Nano Banana," publicly named gemini-2.5-flash-image. The migration is mostly a one-function rename and a response structure update, around 90 minutes of work for most codebases. The catch: one Imagen capability, mask-based editing, has no replacement at all. That separate deadline hits June 30.

What Goes Dark and When

Firebase AI Logic's migration documentation confirms these shutdown dates:

• imagen-3.0-generate-002 — June 24, 2026
• imagen-4.0-generate-001 — June 24, 2026
• imagen-4.0-ultra-generate-001 — June 24, 2026
• imagen-4.0-fast-generate-001 — June 24, 2026
• imagen-3.0-capability-001 (mask editing: inpainting, outpainting, object removal) — June 30, 2026

Vertex AI runs on a slightly different clock — Google recommends migrating before June 30, with a hard shutdown date of August 17 for Vertex AI users on legacy Imagen endpoints. Firebase AI Logic is the shorter deadline. Don't assume extra time if your app uses the Firebase SDK.

The Core Migration: Python

Three things change simultaneously: the method name, the model identifier, and the response structure. All three break if you miss any one of them.

Before (Imagen):

import google.generativeai as genai

client = genai.Client(api_key="YOUR_KEY")

response = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt="A red fox running through snow",
    config={"number_of_images": 1, "output_mime_type": "image/jpeg"}
)

image_bytes = response.generated_images[0].image.image_bytes

After (Gemini Image):

import google.generativeai as genai

client = genai.Client(api_key="YOUR_KEY")

response = client.models.generate_content(
    model="gemini-2.5-flash-image",
    contents="A red fox running through snow"
)

image_bytes = response.candidates[0].content.parts[0].inline_data.data

generate_images becomes generate_content. The response path shifts from .generated_images[0].image.image_bytes to .candidates[0].content.parts[0].inline_data.data. The config dict drops entirely — gemini-2.5-flash-image does not accept number_of_images as a parameter. If you need multiple images, call the API multiple times.

The response structure change is where most migration bugs land. Imagen returned a typed ImageGenerationResponse with a strongly-typed .images array. Gemini Image returns a GenerateContentResponse with a mixed .parts array. If your call generates both text and an image in the same response (Gemini Image supports this; Imagen did not), you cannot reliably grab index 0 — you need to filter parts by type.