Google shuts down the Imagen 4 API on Aug 17 — migrate now

Three imagen-4.0 endpoints go dark Monday; Google points you to gemini-3.1-flash-image, and a bigger retirement wave hits Gemini 2.5 this fall.

Nowline Aug 15 4:00 PM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • Three Imagen 4 endpoints go dark Monday

    On August 17, Google retires imagen-4.0-generate-001, imagen-4.0-ultra-generate-001, and imagen-4.0-fast-generate-001 from the Gemini API — the earliest-possible shutdown for endpoints it deprecated back in June. Any code still calling those model IDs starts failing that day.

  • Where to go: gemini-3.1-flash-image

    Google's named migration target is gemini-3.1-flash-image (aka Nano Banana 2), which runs roughly $0.04–$0.15 per image depending on resolution. There's no migration fee — you're simply on current Gemini API image pricing and quota.

  • It's not a drop-in swap — test first

    Google tells teams to move representative prompts over and re-check aspect ratios, safety filters, SynthID provenance, latency, quota, and cost before flipping. Nano Banana is a conversational image model, so prompt behavior and outputs differ from Imagen 4's.

  • This is the front edge of a bigger Google purge

    More deadlines are stacking: gemini-robotics-er-1.6-preview retires Aug 31, gemini-2.5-flash-image on Oct 2, and the Gemini 2.5 text models (pro, flash, flash-lite) around Oct 16. Published dates are 'earliest possible,' so audit your pinned model IDs now.