Google releases Nano Banana 2.1, based on Gemini 3.6 Flash, saying it improves on previous versions "across the board"; pricing is ~50% lower vs. Nano Banana 2 (Matthias Bastian/The Decoder)

Google released Nano Banana 2.1 on October 6th, 2026, without fanfare. The model appeared in the Google Flow model selector on October 5th, disappeared, and returned the next day with a brief post on @GoogleAIStudio, claiming it "outperforms our previous models in every aspect". Pricing for image generation is the key factor for users.

The Nano Banana 2.1 costs half the price of Nano Banana 2 for all shared resolutions. Google's documentation lists "precise text rendering" as one of its improvements. This guide covers what changed, the cost, limitations, and how to test the model via API.

Specifications:

- Model ID: gemini-nano-banana-2.1

- Output resolutions: 1K, 2K, 4K

- Inputs: Text, image, video without audio

- Reference images: Up to 14

- Thinking: Enabled by default and cannot be disabled on the API

- Intermediate thinking outputs: Up to two, with no additional charges

- Search grounding: Uses Google Web Search and Google Image Search

Improved aspects:

- Better design visual

- Mask-based editing

- Greater subject consistency

- More natural images

- Precise text rendering

- Search grounding with Google Image Search

Text rendering specifics:

- Generates readable, stylized text for infographics, menus, diagrams, and marketing assets

- Precise text rendering is a specific improvement in version 2.1

- This claim is from Google, not a measured result

- No public benchmark of text rendering precision for version 2.1

Testing:

- Test real product prompts before migrating production traffic

- Menu and menu cards

- Product labels

- Posters

- Slides

- Mockups

- Infographics

- Mask-based editing allows modifying specific parts of an image by describing them in the prompt, preserving the rest of the image. This "semantic masking" can be used for menus, labels, posters, slides, and infographics.