Google's Imagen 4 API shuts off today — migrate to Nano Banana now

Three imagen-4.0 endpoints stop returning images at the cutoff — the exit is gemini-3.1-flash-image, with a new call shape and a pricing swing to plan for.

Nowline Aug 17 6:00 AM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • The cutoff is real, and it's today

    Google's changelog pulls imagen-4.0-generate-001, -ultra-generate-001, and -fast-generate-001 on Aug 17. These aren't soft deprecation warnings — calls to those endpoints stop returning images, so any app still wired to them breaks now.

  • Nano Banana is the exit door

    Migrate to gemini-3.1-flash-image. The call shape changes: swap generate_images() for generate_content(), and read images out of response.candidates[0].content.parts instead of a flat array. Legacy gemini-3-image variants are being retired alongside it.

  • The price flips on whether you batch

    Standard-resolution generation runs about $0.067/image against Imagen 4's $0.04 — a 67% jump on the naive path. Route it through the Batch API and it drops to roughly $0.034, actually undercutting the model you're leaving.

  • Audit for the features that vanish

    The replacement drops negative prompts, explicit image-format control, and optional watermarking, and returns one image per call — you loop for variants. Any pipeline leaning on those needs a rethink before you flip the model name.

  • Elsewhere: Cursor's fast Cloud Agents turn on by default

    Cursor's prebuilt Cloud Agent environments — the ones that boot up to 3x faster — switch on by default for all new and existing setups today at no extra cost. Nothing to configure; background agents just start from a warm build.