Gemini 3 Pro Image
Neural Network
Gemini 3 Pro Image is a Google model for image generation and editing, also known as Nano Banana Pro.
Max answer length
(in tokens)
Context size
(in tokens)
Image prompt
(per 1K tokens)
How it works Gemini 3 Pro Image?
Nano Banana Pro, also known as Gemini 3 Pro Image, is a Google model for image generation and editing, built on Gemini 3 Pro. Compared to the first Nano Banana, it features significantly stronger multimodal reasoning and real-world grounding: the model parses complex text descriptions more accurately and makes edits based on references more carefully without breaking the original scene. In BotHub, it is available from Russia without a VPN or foreign card, and you pay as you go, in rubles, for consumed tokens — no subscription required. With over 250 models in the same window, it's easy to compare results with other generators, and you can connect images to your product via a unified OpenAI-compatible API, switching models without rewriting the integration. This is useful for designers creating concepts and posters; for marketers creating ads and social media content, including background, object, and detail replacement in finished shots; for product teams creating interface illustrations and covers; and for content creators producing series of images in a consistent style from a single reference. Companies have access to contracts, invoices, electronic document management (EDO), and an admin panel with limits.Frequently asked questions about Gemini 3 Pro Image
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
nano-banana-pro is a Google model for working with images: it creates pictures from text descriptions and redraws existing ones based on a reference. It is suitable for illustrations, concepts, banners, covers, visuals for social media and presentations, and batch mode helps prepare a series of variants at once.
Specific resolutions and file formats are not recorded in our data, so we won't promise specific numbers. Focus on a detailed description of the shot: composition, lighting, style, proportions — the prompt has the strongest influence on the result. The easiest way to check the outcome is to generate a couple of test variants in BotHub.
A separate reasoning mode before answering is not noted in our data, but that doesn't mean it doesn't exist. In terms of architecture, the model belongs to a family where image generation relies on text parsing, so it takes the detailed scene description into account as a whole.
There are no notes about function calling or strict JSON output in our data, but we won't deny these capabilities. The main scenario here is different: the model outputs an image. If you need a structured response, build the schema using a text model — via the unified BotHub API, this can be done without rewriting the integration.
You can upload images — this is the image-to-image mode: you provide a reference and ask to change the style, background, composition, or assemble a new shot based on it. Audio and video inputs are not noted in our data, so count on a text-plus-image combination.
The language of prompts and output is not fixed in our data, so we won't promise anything specific — test it with your tasks, and make the scene description as detailed as possible. The BotHub interface and support are in Russian, payment is via Russian cards in rubles, and access is available without a VPN or foreign card.