Google's Nano Banana 2.1 Image Model Is Generally Available With 14 Reference Images

Google made Nano Banana 2.1 generally available on 6 October 2026. It is the newest image generation and editing model in the Gemini family, and the API model code is gemini-nano-banana-2.1. The older gemini-3.1-flash-image model, known as Nano Banana 2, is deprecated, and developers are told to migrate.

Google positions it as an update that keeps Flash-level speed and cost while improving visual quality, prompt adherence, character consistency and text rendering. It outputs at 1K, 2K and 4K, and Google says it fixes tiling artefacts on panoramic aspect ratios at 2K and 4K.

The feature that matters for creative work is reference handling. It supports multi-image fusion with up to 14 reference images, with character consistency for up to four characters and object fidelity for up to ten objects. For anyone building a campaign, consistency across a set of images is the hard part, and this is aimed squarely at it.

Google's own preference scores put the thinking variant at 1050, against 990 for Nano Banana 2 with thinking and 935 for Nano Banana Pro. These are internal numbers, not independent tests, so treat them as a direction rather than a verdict.

On price, Google's documentation says images cost $30.00 per million output tokens, which it equates to about $0.113 per 4K image, with batch pricing at roughly half. Check the pricing page for input token costs before you plan a big run, because reference images count towards them.

It is available in the Gemini app, Google AI Studio, the Gemini API, Search AI Mode, Google Ads, Flow and Stitch.

Our view: we have used AI imagery where a photo shoot did not make sense, for example on a dog walking booking platform, and the lesson was that the model gets you a draft and judgement finishes it. Better reference handling makes the draft closer to the brief. It does not replace an art director who can say what is wrong with it.