Nano Banana Pro (Gemini 3 Pro Image)

Nano Banana Pro is an image generation and editing model from Google, also called Gemini 3 Pro Image. It was announced on 20 November 2025 and is built on Gemini 3 Pro, the model Google uses for its general-purpose multimodal work. In practice, that means the picture-making is done by the same underlying model that handles text and other kinds of input, rather than by a separate image-only system.^[1]

Where the name and the versions come from

The previous model in this line circulated under the nickname "Nano Banana"; Google's own description refers to it as the original Nano Banana, and it is what the Gemini app falls back to when a free user's quota runs out.^[1] The new model first appeared to developers as a preview, under the identifier gemini-3-pro-image-preview, released on 20 November 2025. The generally available version, gemini-3-pro-image, followed on 28 May 2026.^[2]

What it can do

The capability Google leads with is text. The model renders text inside an image, from short taglines to long paragraphs, and it can generate or translate text in multiple languages.^[1] A poster, a diagram label or a line of copy in a mock-up no longer has to be faked with letter-shaped noise.

It can also connect to Google Search while generating, pulling in current information to build infographics, diagrams and recipe snapshots.^[1] A picture can therefore carry facts that were not in the prompt, and that were not in the model's training data either.

On the output side it offers 2K and 4K resolutions, and it can blend up to 14 input images in a single result while keeping up to five people looking like themselves.^[1] That combination — many references, several recognisable faces, high resolution — points at composite work such as group shots, product scenes and layouts rather than single-subject illustration.

Where it runs

The model rolled out across a long list of surfaces: the Gemini app, AI Mode in Search (in the US, for Pro and Ultra subscribers), NotebookLM, Google Ads, Slides and Vids, the Gemini API, Google AI Studio, Google Antigravity, Vertex AI and Flow.^[1] NotebookLM's Infographic and Slide Deck outputs, added in November 2025, were powered by Nano Banana Pro.^[3] Free-tier users of the Gemini app receive limited free quotas and fall back to the original Nano Banana model when those run out.^[1]

Checking whether an image is Google-made

Inside the Gemini app, users could upload an image and ask whether Google AI generated it, with the answer based on SynthID.^[1] The feature initially accepted English prompts only, with more languages as well as audio and video planned.^[1]

Where it sits

Taken together, the features move an image model a step away from illustrating a document and towards producing one. Text has to be readable, the information in the picture has to be current, and the resolution has to survive being placed in a layout. That is also why the model was wired into surfaces such as Slides, Vids and Ads, where an image is one element of a page and a misspelled word on it spoils the whole thing.^[1]