Overview
Gemini 3.1 Flash Image, also called Nano Banana 2, is Google’s image-understanding and generation model designed for fast, iterative visual creation. Google announced it on February 26, 2026. (blog.google)
Capabilities and uses
- Generates images from instructions and supports image editing across multiple turns. It can combine text and image output. (docs.cloud.google.com)
- Supports resolutions from 512px to 4K and multiple aspect ratios. Google describes improved instruction following, text rendering and translation, and subject consistency—up to five characters and 14 objects in a workflow. (blog.google)
- Suitable uses include infographics, diagrams from notes, data visualizations, marketing mockups, localized imagery and storyboards. Google Search grounding can supply current information and images when enabled. (blog.google)
Limits and availability
- The specified ID was a public-preview endpoint. Google Cloud lists July 17, 2026 as its deprecation date and recommends migrating to
gemini-3.1-flash-image; check endpoint availability before use. This does not establish the status of the preview ID on every platform. (docs.cloud.google.com) - Google Cloud documents a 131,072-token context window, a 32,768-token output limit and a maximum of 14 input images per prompt. Audio and video are not supported. The documented knowledge cutoff is January 2025. (docs.cloud.google.com)
- The Cloud preview documentation lists no support for structured output, context caching, function calling or model tuning. Generated images and text should still be checked for accuracy and suitability. (docs.cloud.google.com)
Pricing
Checked November 16, 2026. A current price specifically for the deprecated gemini-3.1-flash-image-preview ID could not be verified. Google’s current Gemini Developer API pricing instead lists the replacement gemini-3.1-flash-image; the following are USD standard paid-tier prices for that listed ID, not a verified quote for the deprecated endpoint: (aistudio.google.com)
- Input: $0.50 per 1 million text or image tokens.
- Output: $3 per 1 million text and thinking tokens, or $60 per 1 million image tokens.
- Approximate image-output equivalents: $0.045 per 0.5K image, $0.067 per 1K image, $0.101 per 2K image and $0.151 per 4K image. These are token-based equivalents, not additional charges.
- The listed API free tier is unavailable for model input and output. Search grounding has separate charges and conditions; platform and endpoint pricing may differ. (aistudio.google.com)