Nano Banana 2 Review

The fruit naming has stopped being a joke and started being a product line. Nano Banana 2 is Gemini 3.1 Flash Image, and it took over the Gemini app on February 26, replacing Nano Banana Pro for everyone including free accounts. Google's claim is roughly ninety five per cent of Pro quality at Flash speed and about half the cost, and in practice the gap only shows on the hardest text heavy compositions. Everything else is better: lighting, texture, the small print in a mockup, and the thing image models used to fail at, which is keeping the same face across a dozen edits. It tracks up to five characters and fourteen objects, at anything from 512 pixels to 4K.

Where it sits matters more than how good it is. This is now the default image model in the AI Mode of Search and in Lens, in Flow, in Google Ads, in AI Studio and in the Gemini API, and it is free in the app. Paying subscribers can still fall back to Nano Banana Pro from the regenerate menu, which is the right way round: the fast model is the floor rather than the compromise.

The caveats are the ones this whole category carries. It does not reason about a brief the way a person would, so it will follow a bad prompt off a cliff, and the thinking and coding grades below record a tool that does neither and claims neither. Every output carries SynthID and C2PA content credentials, which is welcome and is also not a solution to anything on its own. The price is the strongest argument in its favour: free in the Gemini app, and around fifteen cents an image for 4K output on the paid API.