Nano Banana Pro: Gemini 3 Image Features, Price & Access

What shipped

Google's announcement describes an image model built on Gemini 3 that blends up to 14 reference images while holding the likeness of up to 5 people consistent across a scene, renders legible text in multiple languages and lets a user re-render that text or translate it, and outputs at 1K, 2K or 4K resolution. Google also markets it as able to pull in Google Search results while composing an image, for cases like visualizing current data or a recipe. [2]

The rollout was wide and simultaneous rather than staged: TechCrunch's coverage lists the model going live the same day as the default image generator in the Gemini app, in AI Mode in Search for Pro and Ultra subscribers, in NotebookLM's paid tiers, in Google Slides and Vids for Workspace customers, in the Flow video tool for Ultra subscribers, and in the Gemini API, Google AI Studio and the Antigravity IDE for developers. [3]

Pricing stepped up sharply with the upgrade: TechCrunch reports the original Nano Banana priced at $0.039 per 1024px image, against $0.139 per 1080p or 2K image and $0.24 per 4K image for the Pro version, an API cost roughly 3 to 6 times higher depending on the resolution requested. [3]

An image model gated by a language model

Atlas interpretation: The event's title, "Google puts Gemini 3 Pro behind an image model," points at the arrangement rather than the feature list: Nano Banana Pro is not a standalone diffusion model wearing a new name, it is Gemini 3 Pro's reasoning wrapped around image generation. Gemini 3 Pro itself had shipped only two days earlier, on November 18, 2025, and Nano Banana Pro is the first consumer-visible product built on top of it rather than a chat interface to it. The multi-step edits Google demonstrates, like keeping a diagram's labels consistent while re-rendering it in another language, depend on that underlying reasoning, which is also what the review models charge for at 3 to 6 times the prior price. [2][3]

Atlas interpretation: The name carries its own history. The original Nano Banana spent weeks tested anonymously on LMArena under that codename before Google confirmed it as Gemini 2.5 Flash Image, and the nickname stuck hard enough that Google kept it as the public-facing brand for the Pro follow-up rather than leading with the model number. Google's own model card for the new release still lists it as "Nano Banana Pro," with "Gemini 3 Pro Image" as the parenthetical. [2][4]

Provenance and verification

Google states that every image the model produces carries an imperceptible SynthID watermark, and that a user can upload an image to the Gemini app to check whether it was generated or edited by Google's AI. Google's own model page repeats the SynthID claim without giving technical detail on the watermark's detection accuracy or robustness to common edits like cropping or recompression. [2][4]