Make image creation and editing a conversation: describe a change in words and get an accurate, consistent picture, cheaply enough to run inside Search, Gemini and developer apps at huge scale.
Nano Banana 2.1 went generally available Oct 6, 2026; the Nano Banana 2 API model shuts down Oct 29.
Updated 5d agoChecked 11 Oct1 update this week
Image output is built into a Gemini language model, so the same network that understands the request and the input pictures also produces the new image.
You can keep asking for changes in follow-up turns; the model carries the earlier images forward, which keeps faces and objects the same between edits.
It can run Google web and image searches before drawing, so maps, products or recent events in a picture match real references instead of guesses.
Flash models trade a little control for low cost and speed; Pro spends more compute per image for fine lighting, camera and layout control.
| Spec | Nano Banana | Nano Banana Pro (Gemini 3 Pro Image) | Change |
|---|---|---|---|
| Max output resolution | 4KR (reported) | 1024 x 1024 (original Nano Banana) | —no comparable change |
| Consistent characters (Nano Banana 2) | Up to 5R (reported) | — | —no comparable change |
| Consistent objects (Nano Banana 2) | Up to 14R (reported) | — | —no comparable change |
| Reference images (2.1) | Up to 14R (reported) | — | —no comparable change |
| Input context | 131,072 tokensR (reported) | — | —no comparable change |
| Price per 1,000 images (NB2 launch) | About $67R (reported) | About $134 (Nano Banana Pro) | —no comparable change |
R reported by the company
Google DeepMind
Gemini, AlphaFold, Genie, Veo, TPU chips
Google DeepMind is Alphabet's AI lab, formed in 2023 from DeepMind and Google Brain. It builds the Gemini models and leads in AI for science. AlphaFold won a share of the 2024 Nobel Prize in Chemistry.
