Imagen 4 API Shutdown: How to Migrate Your Image Generation

Imagen 4 API shutdown at a glance

Google's deprecations page, last updated 2026-09-15 UTC, still lists August 17, 2026 as the shutdown date for those three GA IDs. Preview Imagen 4 IDs imagen-4.0-generate-preview-06-06 and imagen-4.0-ultra-generate-preview-06-06 had earlier shutdown dates and are a different cutover.

Which Imagen 4 model IDs were shut down?

The June 15, 2026 changelog named these generally available Imagen 4 Gemini API IDs:

• imagen-4.0-generate-001
• imagen-4.0-ultra-generate-001
• imagen-4.0-fast-generate-001

Do not treat older preview IDs, Imagen 3 IDs, or Gemini native image preview IDs as part of this same list. Those families carry their own dates on the deprecations table.

What replaces Imagen 4?

There is no single drop-in rename. Google's surfaces still disagree on how narrowly they describe the successor, so this page separates the shutdown fact from the current replacement choices.

• The Gemini deprecations table names gemini-3.1-flash-image as the recommended replacement for all three GA Imagen 4 IDs.
• The current image-generation documentation (updated 2026-09-04) centers the Nano Banana family and says Imagen models are deprecated and shut down on August 17, 2026. It describes Nano Banana 2 as the versatile general-purpose image model and Nano Banana 2 Lite as the fastest and cheapest option.
• The Gemini pricing page checked on 2026-09-15 no longer lists Imagen 4 prices or an Imagen 4 shutdown warning. It publishes token-based Nano Banana rates and separately warns that gemini-2.5-flash-image shuts down on October 2, 2026.

Use the current IDs below as a menu, not as a promise that every Imagen 4 job maps to one of them without code or quality changes:

Prefer gemini-3.1-flash-image when you want the ID Google currently lists as the Imagen 4 replacement. Choose Lite when latency and cost dominate and the job does not need multi-reference editing. Choose Pro when the live docs for that job still call for the premium Gemini image model. Do not treat legacy gemini-2.5-flash-image as a durable Imagen 4 successor — Google already dates its own shutdown.

Imagen 4 to Gemini image migration checklist

1. Search application code, infrastructure, and model-fallback lists for all three GA Imagen 4 IDs.
2. Map each retired ID to the workloads that still depend on it (generation, editing, batch, or fallback).
3. Choose a currently supported Gemini image model from live Google docs, not from a stale blog post.
4. Update the request surface — Interactions or generateContent — instead of swapping only the model string.
5. Retest prompt fidelity, aspect ratio, and output dimensions on a production-like sample.
6. Retest image-editing and reference-image flows as their own suite. Do not assume text-to-image success covers them.
7. Validate error handling, quotas, and rate-limit behavior on the new endpoint.
8. Compare cost on the same production mix. Gemini image models are token-priced; Imagen 4 rates are no longer published.
9. Run side-by-side visual QA before shifting traffic.
10. Cut over production, then monitor failures, latency, and spend. Remove retired IDs from every fallback path.

Code migration examples

These Gemini examples follow the current image-generation documentation. They are example targets, not guaranteed one-for-one replacements for every Imagen workload. Confirm the SDK version and field names against Google's live docs before you ship.

Python

from google import genai
import base64

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-3.1-flash-image",
    input="Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme",
)

with open("generated_image.png", "wb") as f:
    f.write(base64.b64decode(interaction.output_image.data))

JavaScript

import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";

const ai = new GoogleGenAI({});

const interaction = await ai.interactions.create({
  model: "gemini-3.1-flash-image",
  input:
    "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme",
});

if (interaction.output_image) {
  const buffer = Buffer.from(interaction.output_image.data, "base64");
  fs.writeFileSync("gemini-native-image.png", buffer);
}

REST

curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.1-flash-image",
    "input": [
      {
        "type": "text",
        "text": "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme"
      }
    ]
  }'

OfflineCreator Studio CLI

Studio is not a Gemini API proxy and does not accept imagen-4.* or gemini-*-image IDs. If you want hosted generation through Studio's catalog instead of Google's Gemini API, the published CLI starts a generation with a Studio model ID:

npx @offlinecreator/mcp generate \
  --model nano-banana-2-lite \
  --prompt "A still life of citrus on dark slate, studio lighting" \
  --aspect 1:1

That command uses the Studio ID nano-banana-2-lite, which maps to the managed endpoint google/nano-banana-2-lite and currently quotes 6 credits at the catalog default. It is not a drop-in Gemini Interactions call. See the CLI generate-image guide and the Studio model identifier page.

Imagen 4 vs current Gemini image models

Values below were checked against Google's public docs on 2026-09-15. Pricing cells use Google's published paid-tier 1K image equivalents where Google states them. Imagen 4 prices are omitted because they are no longer on the current pricing page.

Migration mistakes to avoid

Reduce future model-ID churn in application code

Hard-coding one vendor image ID is what made this shutdown expensive. OfflineCreator Studio is a hosted credit catalog with web, MCP, CLI, and REST surfaces over modelId values from the current Studio catalog. It does not claim to route Imagen 4, Nano Banana 2 (gemini-3.1-flash-image), or arbitrary Gemini image IDs.

What this repository does verify today: the launch catalog includes Nano Banana 2 Lite (nano-banana-2-lite → google/nano-banana-2-lite) and Nano Banana Pro (nano-banana-pro → fal-ai/nano-banana-pro). Those jobs run through Studio's managed fal path. Credit quotes come from published Studio pricing, not from Google's token card.

If you still need Google's Gemini API contract — Interactions, grounding, or an ID Studio does not catalog — keep that integration on Google and use this page as the migration checklist. If you want a hosted catalog plus MCP/CLI instead of another provider-specific client, start at the AI image generator, compare models on AI models, or read API vs MCP vs CLI and migrate a provider script to Studio REST.

Teams whose actual goal is to stop using hosted image APIs can look at LocalForge on offlinecreator.com. That is a local/offline product, not a Studio Imagen 4 shim, and it is not this page's primary path.

FAQ