Google Is Retiring Imagen 4 — If Your Image Generation Runs on Antbase, You're Already Migrated
·4 min read
🎨

Google Is Retiring Imagen 4 — If Your Image Generation Runs on Antbase, You're Already Migrated

Google is discontinuing the Imagen 4 endpoints on Aug 17, 2026, pointing everyone to Gemini 3.1 Flash Image. If you call Imagen directly you have to update code before then. On Antbase, image generation routes through ant:image-gen — already on the newer Gemini models, plus OpenAI's, behind one API.

AntbaseGoogleImagenGeminiimage generationmodel deprecationmigration

Google just sent a deprecation notice for image generation: the Imagen 4 endpoints are being shut off on August 17, 2026. After that, calls to them return a 404. The recommended migration path for all three is the same — Gemini 3.1 Flash Image.

  • •imagen-4.0-generate-001 -> gemini-3.1-flash-image
  • •imagen-4.0-ultra-generate-001 -> gemini-3.1-flash-image
  • •imagen-4.0-fast-generate-001 -> gemini-3.1-flash-image

If you call Imagen directly, that's a code change you have to make before the deadline. If your image generation runs through Antbase, there's nothing to do.

Antbase routes image generation, not just text

Antbase has a virtual model for this — ant:image-gen — that routes across every image-capable model in the catalog. It's already pointed at the newer Gemini image models (gemini-3.1-flash-image, gemini-2.5-flash-image, gemini-3-pro-image) — the exact migration target Google is recommending — alongside OpenAI's gpt-5-image. The legacy Imagen 4 endpoints were never in the routing path, so the August shutoff changes nothing for you.

python
# Was: a pinned Google endpoint that's now on a kill date
#   model = "imagen-4.0-generate-001"

# Now: one stable handle that routes to the best live image model —
# already the Gemini 3.x family, with OpenAI's as an alternative.
model = "ant:image-gen"
# or pin a specific one: "google/gemini-3.1-flash-image-preview"

Image models churn exactly like text models

Imagen 4 to Gemini 3.1 is the same story we just saw with OpenAI retiring GPT-5 snapshots: vendors version, deprecate, and retire models on their own clock, and whoever pinned the old ID is the one stuck migrating. The fix is the same too — don't pin a single model. Route through a stable handle and let the deprecations be someone else's problem.

Compare the image models Antbase routes to, side by side, in the playground:

One note: this is about Antbase's own routing. If you have other apps calling Imagen 4 directly on your Google project, those still need updating before August 17 — but anything you move behind ant:image-gen is migration-proof going forward.

Try ANT routing today

Drop-in OpenAI-compatible API. Change your base URL and every request gets intelligent routing across 30+ providers.