Google has released Nano Banana 2.1, an image generation and editing model built on Gemini 3.6 Flash that can also emit accompanying text. The update raises the input context to 1 million tokens (with output buffers of 4K tokens for graphics and 64K for text), so large text descriptions and reference image sets can be handled in a single request. Google reports that the model beats both Nano Banana 2 and Nano Banana Pro on generation and editing benchmarks, reaching 1050 Elo in preference ratings in reasoning mode, with the biggest gains in character and object consistency across frames, mask and sketch editing, and factual accuracy in infographics drawn from Gemini’s knowledge base. Nano Banana 2.1 is available in the Gemini app, AI Studio, the API, Flow, and Stitch.

Nano Banana 2.1 model card

Related: Google Unveils Nano Banana 2 Lite and Gemini Omni Flash, Google Expands Gemini Family With 3.6 Flash, 3.5 Flash-Lite, and Cyber-Security Model