WHY THIS MATTERS
Google is shutting down three Imagen 4 family models available via API this Monday, August 17, 2026, ending support for imagen-4.0-generate-001, imagen-4.0-ultra-generate-001 and imagen-4.0-fast-generate-001. Anyone using those models needs to migrate to gemini-3.1-flash-image, known as Nano Banana 2, but the switch is not a simple swap: the generate_images() method no longer exists, and image generation now runs inside generate_content(), requiring code rewrites, fresh aspect ratio testing and a higher per image price than before.In this article
Three Imagen 4 Models Go Offline This Monday
Google is ending API support this Monday, August 17, 2026, for three identifiers in the Imagen 4 family: imagen-4.0-generate-001, imagen-4.0-ultra-generate-001 and imagen-4.0-fast-generate-001. The information appears on Google's own deprecations page and was detailed by outlets including Kingy AI and Byteiota, which track the company's model retirement schedule. As of today, API calls still pointing at those three model IDs will start failing.
The Official Replacement: Gemini 3.1 Flash Image, aka Nano Banana 2
Google is steering developers toward gemini-3.1-flash-image, nicknamed internally and across the market as Nano Banana 2, as the image generation successor within the Gemini family. It happens to be the very same model MaxAssistant's own editorial team uses to produce the cover images for this site's articles, which makes the change a direct, practical matter for anyone already relying on AI image generation day to day.
Not a Drop-In Replacement: the Old Method Simply Disappeared
The migration is far from automatic. According to reporting from Byteiota and Narracomm, the generate_images() method used to call Imagen 4 has been completely removed from the SDK: image generation now happens inside the generate_content() method, the same one used for text, with a different API call format. Teams that integrated Imagen 4 directly into their products need to rewrite chunks of code, not just swap a model name in a variable.
What Needs Retesting Before Pipelines Break
Beyond the method change, teams that relied on Imagen 4 need to revalidate several behaviors that may have shifted in Gemini 3.1 Flash Image: prompt adherence, supported aspect ratios, how the SynthID watermark automatically applied to images is handled, response latency, and account quota limits. Skipping this retesting step is the most common way production pipelines quietly stop working without warning.
Prices Rise: From Up to $0.06 to $0.067 Per Image
The math changes too: a 1K image on Gemini 3.1 Flash Image costs $0.067, compared to a range of $0.02 to $0.06 charged by Imagen 4, depending on the model version used. For products generating large volumes of AI images, such as e-commerce catalogs, ad creative at scale, or editorial covers like the ones on this site, that per unit price increase adds up quickly in monthly infrastructure costs.
Why It Matters for Brazilian Agencies and SMBs
Any Brazilian agency or company that automated image generation through Google's API, whether for marketing assets, product mockups or editorial content, needs to audit today whether any workflow still calls the three deactivated Imagen 4 IDs. The practical recommendation is simple: search the codebase for references to imagen-4.0-generate-001, imagen-4.0-ultra-generate-001 or imagen-4.0-fast-generate-001, migrate those calls to generate_content() with the gemini-3.1-flash-image model, run a batch of tests comparing image quality and aspect ratio before and after, and recalculate monthly cost factoring in the higher per image price. Ignoring this warning means a real risk of production breaking starting today.