Models·Google

Nano Banana Pro (Gemini 3 Pro Image Preview)

google/gemini-3-pro-image-preview

Google’s image-generation and editing model built on Gemini 3 Pro, with text and image input.

$2Input, per million tokens
$12Output, per million tokens
66KContext window
33KLongest answer

66K tokens of context — about 49,152 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Vision
  • Takes image
  • Takes text
  • Gives image
  • Gives text

Released 20 Nov 2025 · checked 3 Oct 2026.

About it

Nano Banana Pro, also called Gemini 3 Pro Image Preview, is a Google model for generating and editing images. It accepts text and images, and returns images or text. It supports multimodal reasoning and has a 65,536-token context window. Input costs $2 per million tokens, while output costs $12 per million tokens.

Review

In a line

Nano Banana Pro is Google’s image-generation and editing model built on Gemini 3 Pro. Its model ID is google/gemini-3-pro-image-preview.

What it's good at

It suits image creation and editing tasks that combine words with pictures. You can provide text, an image, or both, and receive an image or text.

The model has vision and reasoning capabilities. Google describes it as an advance over the original Nano Banana in multimodal reasoning and real-world grounding.

Price in plain terms

Input costs $2 per million tokens. Output costs $12 per million tokens, so generated responses cost more than text sent to the model.

As rough examples, reading a 500-word page costs about 0.13 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.8 cents. 1,000 such turns cost about $8.00.

Context and limits

The context window is 65,536 tokens. That is the amount of text and other input the model can consider in one request.

A single answer can contain up to 32,768 tokens.

Ways to use it

Use it by sending text instructions, images, or both for image-generation and editing work. It can return an image or a text response.

For a broader comparison, see image generation models and Google models.

Verdict

Choose Nano Banana Pro when you need one model to understand pictures and text while creating or changing images. Its large context window helps with long inputs, and its output can be either an image or text.

Skip it if you need a stated price below its $12-per-million output rate.

Verdict

Nano Banana Pro is a fit for people comparing image models by token cost and for workflows that mix written instructions with pictures. It offers both image and text responses, plus reasoning and vision capabilities. The pricing is easy to estimate for text-heavy use, but output tokens are relatively costly compared with input tokens.

  • Accepts both text and image inputs
  • Can return images or text
  • Includes vision and reasoning capabilities
  • 65,536-token context window
  • Output tokens cost more than input tokens
  • Output costs $12 per million tokens
  • Pick it ifImage generation and editing with text or picture inputs; Readers comparing model costs per million tokens; Tasks needing vision and multimodal reasoning
  • Skip it ifYou need a lower output-token price

Pricing

Input is $2 per million tokens and output is $12 per million tokens. Roughly, one 500-word page costs about 0.13 cents to read, one typical chat turn costs about 0.8 cents, and 1,000 such turns cost about $8.00.

Questions

What is Nano Banana Pro?

It is Google’s image-generation and editing model built on Gemini 3 Pro. Its model ID is google/gemini-3-pro-image-preview.

What can I send to the model?

You can send text, images, or both.

What can it produce?

It can return images or text.

How much does input cost?

Input costs $2 per million tokens.

How much does output cost?

Output costs $12 per million tokens.

How much context can it handle?

Its context window is 65,536 tokens, and its maximum output is 32,768 tokens.

When was it released?

The listed release date is November 20, 2025.

On these boards

More from Google