Nano Banana 2.1 API
Google DeepMind's efficient image generation and editing model built on Gemini 3.6 Flash, with 4K output, masked local edits, and multi-reference subject consistency.
Nano Banana 2.1 is Google DeepMind's image generation and editing model, released in October 2026 as part of the Gemini 3 family and built on Gemini 3.6 Flash. It succeeds Nano Banana 2 and serves as the efficient tier below Nano Banana Pro, keeping Flash-level speed and cost while pushing image quality close to the Pro model.
The model is strongest at finished visual design: typography, whitespace, hierarchy and multi-element layout for posters, UI mockups and infographics. It supports mask- and doodle-guided local edits that leave the surrounding composition intact, accepts up to 14 reference images to hold 4 characters and 10 objects consistent across images, and renders legible text on thumbnails, logos, product labels and infographics. Output is available at 1K, 2K or 4K, with reduced tiling artifacts on extreme aspect ratios. Text, image, video and PDF are accepted as input; text and image are returned. Grounding with Google web and image search, multi-turn conversational editing, and configurable thinking levels (minimal, medium, high; default medium) are supported.
It fits design workflows that need professional precision with fast iteration: infographics, ad creative, product photography, character design and UI work. Known limitations include imperfect character consistency between input and output, mask or doodle edits that can leave ink residue or miss instructions, pose alignment that occasionally carries over unintentionally, and errors in counting, hands, strict exclusion instructions and left-right spatial reasoning.
Updated:
MODEL ID
gemini-nano-banana-2.1- Context
- 65.536K
- Max output
- 58.982K
- Input
- image / text
- Output
- image / text
Providers
Compare API prices across providers. Select a provider to view all charges and conditions.
| $1.5 | $7.5 |
Reasoning
- Default effort
- minimal
- Enabled by default
- Yes
- Always enabled
- No
- Available effort levels
- high / minimal / none
API reference
Model-specific protocol details are not published. Check the API reference for integration requirements.
API reference →Supported parameters
- include_reasoning
- max_tokens
- reasoning
- reasoning_effort
- response_format
- seed
- structured_outputs
- temperature
- tool_choice
- tools
- top_p
Frequently asked questions
What is Nano Banana 2.1?
Nano Banana 2.1 is Google DeepMind's image generation and editing model, released in October 2026 as part of the Gemini 3 family and built on Gemini 3.6 Flash. It succeeds Nano Banana 2 and serves as the efficient tier below Nano Banana Pro, keeping Flash-level speed and cost while pushing image quality close to the Pro model. The model is strongest at finished visual design: typography, whitespace, hierarchy and multi-element layout for posters, UI mockups and infographics. It supports mask- and doodle-guided local edits that leave the surrounding composition intact, accepts up to 14 reference images to hold 4 characters and 10 objects consistent across images, and renders legible text on thumbnails, logos, product labels and infographics. Output is available at 1K, 2K or 4K, with reduced tiling artifacts on extreme aspect ratios. Text, image, video and PDF are accepted as input; text and image are returned. Grounding with Google web and image search, multi-turn conversational editing, and configurable thinking levels (minimal, medium, high; default medium) are supported. It fits design workflows that need professional precision with fast iteration: infographics, ad creative, product photography, character design and UI work. Known limitations include imperfect character consistency between input and output, mask or doodle edits that can leave ink residue or miss instructions, pose alignment that occasionally carries over unintentionally, and errors in counting, hands, strict exclusion instructions and left-right spatial reasoning.
How is usage billed?
Your bill depends on the provider you use, its prices and your usage. Longer inputs or other conditions may have different prices.
