Real vs Rendered

Nano Banana Pro

Google Gemini logoBy Google · Google DeepMind (Gemini team)

Google's top-of-the-range image model, built on Gemini 3 Pro for work where precision matters more than speed.

Not enough plays yet — the fool rate appears once this model has been shown in a finished puzzle.

Nano Banana Pro is the nickname for Gemini 3 Pro Image, the most capable image model from Google DeepMind. It launched in November 2025, a few months after the original Nano Banana became a worldwide hit (the Nano Banana 2 page tells the story of the name). Where the other Nano Bananas are built on Google's fast Flash models, Pro is built on Gemini 3 Pro, its larger reasoning model, and Google pitches it for studio-quality work with fine control.

What it does

Because it reasons about a request before drawing, and can check Google Search for up-to-date facts such as the weather or a recipe, Pro is good at pictures that have to be correct as well as attractive: infographics, labelled diagrams, posters and product mock-ups. It writes clear text in many languages and can translate the words inside an existing image while keeping the design. It can blend up to fourteen reference images and keep up to five people looking like themselves, and it lets you change the camera angle, focus, lighting or colour grading, or turn a daytime scene into night. Images come out at up to 4K.

Where you'll see it

At launch it was built into the Gemini app, AI Mode in Google Search for paying US users, NotebookLM, Google Ads, Slides and Vids, and the Flow filmmaking tool, as well as Google's services for developers and businesses. It was the Gemini app's main image model until Nano Banana 2 took over as the default in 2026, and it remains the option for people who want the highest quality and can wait a little longer.

Controversy

Pro's realism quickly raised alarm. Within days of launch, a user in India showed it producing convincing copies of Aadhaar and PAN identity cards for a made-up person, prompting warnings that photo-based ID checks could be fooled. In December 2025 the misinformation watchdog NewsGuard reported that it had produced convincing images for all thirty false claims it tested, from fake news broadcasts to bogus health scares, and that unlike some rivals it did not refuse to show real public figures.

Watermarks

Every image carries Google's invisible SynthID watermark, and Pro launched alongside a feature that lets you upload a picture to the Gemini app and ask whether Google AI made it. Images made by free and Pro subscribers also get a visible mark, but Ultra subscribers and developers can export them without one, and NewsGuard pointed out that the visible mark is easy to crop off.

Sources: Google DeepMind: Nano Banana Pro · OpenRouter: Nano Banana Pro · Business Today · AlternativeTo · NewsGuard: Google's new AI image generator is a misinformation superspreader · Storyboard18

Logo: Google Gemini.

Last reviewed 2026-10-07.

← All models · Play today's puzzle