Nano Banana is the name that Google DeepMind gives to the native image generation of Gemini, Google's multimodal AI model. Multimodal means that one model reads and writes several kinds of data; here it takes text, images or video as input and returns images. The first model with the name, Gemini 2.5 Flash Image, is now the legacy member of a family that also includes Nano Banana 2 (Gemini 3.1 Flash Image), Nano Banana 2.1, Nano Banana 2 Lite and Nano Banana Pro (Gemini 3 Pro Image).
A user describes a picture in everyday language, or uploads images with instructions, and keeps the conversation going to refine the result. The newer models render legible text inside images and can use Google Search for real-world facts. Every generated image carries a SynthID watermark, a mark embedded in the image that identifies it as AI-generated.
The models are offered in the Gemini app, in Google AI Studio and, for developers, through the Gemini API with pay-per-use prices. Google deprecated Gemini 2.5 Flash Image and set its shutdown for 2 October 2026.