Google: Nano Banana 2.1

Google: Nano Banana 2.1 (Gemini Nano Banana 2.1) is an upgraded multi-modal inference image model designed for production-level image generation and editing. It features greater prompt following capabilities, multi-reference editing capabilities, and higher visual fidelity. The model understands text, image, video, and PDF inputs, supports alternate generation of text and images, and can reason about complex spatial and compositional instructions before generating images. Additionally, the model supports basic validation through Google Search and Google Image Search, enabling better long-tail entity recognition, visual understanding, and output accuracy.

By:

Google

Input:

Output:

Context length:

128K

Max output:

33K

Published:

2026-10-06