Imagen 4

What Imagen 4 is

Imagen 4 is a text-to-image model from Google: a user describes a picture in words, and the model generates the image. Google announced it at its I/O developer conference on 20 May 2025.^[1] It belongs to the Imagen family, Google's line of image-generation models, and succeeded Imagen 3.

What Google emphasised

In announcing the model, Google pointed to fine detail — the texture of fabric, water droplets, animal fur — rendered in both photorealistic and abstract styles.^[1] The company also highlighted better spelling and typography.^[1] Text inside generated images had long been a weak point for image models: signs, labels and posters often came out with garbled or invented letters, which limited how useful the output was for design work.

Imagen 4 outputs images at up to 2K resolution, which Google described as better suited to printing or presentations.^[1]

Where it shipped

At launch the model was made available in the Gemini app, in Whisk, on Vertex AI, and inside Workspace applications including Slides, Vids and Docs.^[1] In practice that meant the same model could be reached from a general chatbot, from a dedicated image tool, from Google Cloud's platform for building with AI models, and from the productivity software people use to make slide decks, videos and documents.

Google also announced a faster variant, Imagen 4 Fast, which it said would be up to 10 times faster than Imagen 3.^[1]

The API models and their retirement

Three models became generally available through the API on 24 June 2025: imagen-4.0-generate-001, imagen-4.0-ultra-generate-001 and imagen-4.0-fast-generate-001.^[2] The naming reflects three tiers of the same generation — a standard model, an "ultra" model and a fast model — so that developers could trade off quality against speed and cost when building applications.

Those three models were shut down on 17 August 2026, and Google recommended its Nano Banana image models in their place.^[2] Google's Gemini API deprecation page notes that published shutdown dates are the earliest possible dates, with formal dates announced in advance.^[2] The retirement applied to the API models; the announcement itself described availability in Google's consumer and workspace products.^[2]

Where it sits in the line

Imagen 4 is best understood as a step in a fast-moving sequence. Its improvements were incremental and specific rather than a change of kind: sharper small details, legible lettering, higher output resolution, and a speed-focused variant for high-volume use.^[1] Its lifespan as a publicly offered API model was short — roughly fourteen months passed between general availability and shutdown — which is typical of the pace at which image models have been replaced.^[2] For users, the practical legacy is the feature set it normalised: 2K output and readable text in generated images, carried forward into whatever model Google offers next.