Google Imagen 3

Google DeepMind's image generation page for this entry now resolves to Gemini Image, the model family that has taken over text to image work at DeepMind and is marketed under the Nano Banana name. These models generate and edit images from written prompts, accept uploaded images alongside text instructions, and hold a conversation so a result can be refined over several turns rather than regenerated from scratch. The family targets both casual creators and developers building image features. Three models are published. The pro tier, built on Gemini 3, aims at studio quality precision and control. A Flash tier offers professional level generation and editing at higher speed. A lighter Flash variant is positioned as the fastest and most efficient of the three, built for high volume work at the lowest cost in the lineup. Because the models are built on Gemini, DeepMind describes them as drawing on multimodal understanding and real world knowledge, so generated scenes follow everyday logic. A published prompt guide explains how detail in a prompt maps to control over the output. Google DeepMind does not publish prices on these pages. The models can be tried in the Gemini app and in Google Pics, and built with through Google AI Studio, the Gemini API and the Gemini Enterprise agent platform, where usage is billed under Google's separate API rates. Worth knowing before choosing: this listing is named for an earlier Imagen generation, and the vendor page it pointed to now redirects to the Gemini Image family, so the specific model named here is no longer the one DeepMind presents.