If your app generates images through Google’s API and you haven’t touched that code in a while, today is the day it may have quietly broken. What Happened Today Google retired all three stabl
If your app generates images through Google’s API and you haven’t touched that code in a while, today is the day it may have quietly broken.
Google retired all three stable Imagen 4 endpoints on the Gemini Developer API — imagen-4.0-generate-001, imagen-4.0-ultra-generate-001, and imagen-4.0-fast-generate-001 — effective August 17, 2026. Google’s own deprecation documentation frames listed dates as the earliest possible shutdown time rather than a guaranteed exact hour, but the company has been direct about the end state: once a model is shut down, it is completely turned off, and calls to the retired endpoints will fail outright rather than degrade gracefully.
The recommended replacement is gemini-3.1-flash-image, a model also known internally as “Nano Banana 2” — Google’s general-purpose image generation model built for new Gemini API workloads, offering conversational editing capabilities and low latency at a mainstream price point.
The core technical complication is that Imagen and Gemini image models don’t share an API shape. Imagen used a dedicated image-generation endpoint and a distinct method, `generate_images()`, that returned a clean array of generated images. That method doesn’t exist on Gemini image models at all. Instead, requests go through the more general `generateContent` method, and images arrive embedded as content parts within the response rather than a dedicated image array — developers now have to check each response part for inline image data, and handle cases where the model returns accompanying text alongside the image, something Imagen’s endpoint never did.
A second breaking change: Imagen’s `number_of_images` parameter, which let a single call request multiple image variations at once, has no equivalent on Gemini image models. Each Gemini image call now produces exactly one image, meaning any pipeline that requested batches in a single request needs to be restructured into multiple calls.
Imagen 4 priced output by model tier — a flat rate per image depending on whether you used the standard, fast, or ultra version. Gemini 3.1 Flash Image instead prices by output tokens, which scale with resolution: Google lists standard rates around $0.045 for 0.5K resolution images, climbing to roughly $0.151 for 4K. At standard resolution, that reportedly works out to about 67% more expensive per image than Imagen 4 Standard was — though Google’s Batch API reportedly brings costs down to around $0.034, actually undercutting the old Imagen pricing for teams that can batch their requests rather than needing real-time generation.
This shutdown isn’t an isolated event — it’s one entry in a broader model retirement wave Google has been running throughout 2026 as it consolidates developers onto its newer Gemini 3 model family. The company has been steadily retiring older Gemini 2.5 and earlier Imagen and embedding models on a rolling schedule throughout the year, with Gemini 2.5’s remaining text models slated to follow in October. For teams building on Google’s AI stack, that pattern has practical implications: pinning to a specific model version buys short-term stability but virtually guarantees a forced migration eventually, since Google has shown no indication it plans to support older model generations indefinitely.
This kind of infrastructure churn is happening alongside a genuinely fast release cadence at the top of Google’s model lineup — the company shipped Gemini 3.7 Flash just days ago, three weeks after its predecessor. See our coverage of that release for how quickly Google’s flagship models are iterating even as older endpoints get retired in parallel.
Any team still calling the old Imagen 4 endpoints should expect hard failures rather than a grace period at this point, given Google’s stated policy on shutdown dates. More broadly, developers building on any Google AI API should treat this as a reminder to audit which specific model IDs their production code depends on and check Google’s deprecations table directly, since the company’s retirement cadence shows no signs of slowing down.
Sources: Google AI for Developers, ByteIota, Bez-Kabli
Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.
The post Google Just Shut Off Imagen 4 Today — And the Replacement Isn’t a Simple Swap appeared first on Times Tabloid.
