Nano Banana 2.1: Official Model Profile and Usage Guide
Nano Banana 2.1 is the latest high-efficiency image generation and conversational editing model from Google, designed to deliver significant improvements in visual quality, prompt adherence, and text rendering while maintaining Flash-level speed and cost efficiency. As an update to Nano Banana 2, it serves as the primary workhorse model for image generation tasks, offering enhanced capabilities across multiple resolutions and aspect ratios.
Official documentation
Nano Banana 2.1: Technical Overview
This section covers the core identity of this model, its position within the Gemini model family, and its technical specifications for input and output processing.
Model Identity and Key Updates
This model (gemini-nano-banana-2.1) is Google's latest high-efficiency image generation and conversational editing model. It represents a significant update to Nano Banana 2, maintaining Flash-level speed and cost efficiency while delivering notable improvements across several key areas. The model enhances visual quality and realism across 1K, 2K, and 4K output resolutions, with 1K being the default. A critical fix addresses tiling artifacts on wide and panoramic aspect ratios, specifically 1:4, 4:1, 1:8, and 8:1, at 2K and 4K resolutions. Text rendering and infographic layout accuracy have also been improved, making the model more suitable for content that combines visuals with text elements. These updates position this model as the recommended choice for new projects requiring efficient yet high-quality image generation.
Input and Output Specifications
The model accepts a variety of input data types, including text, image, video, and PDF, and produces both image and text outputs. It features a substantial input token limit of 131,072 tokens and an output token limit of 32,768 tokens, allowing for complex prompts and detailed generations. The model supports multi-image fusion with up to 14 reference images, enabling character consistency for up to 4 characters and object fidelity for up to 10 objects. This makes it suitable for tasks requiring consistent visual elements across multiple generations. The model code is gemini-nano-banana-2.1, and it is documented as a stable version with the latest update in October 2026.