Google DeepMind Unveils Nano Banana 2.1, New Gemini 3 Image Model That Outperforms Nano Banana Pro in Character Consistency

Google DeepMind has unveiled Nano Banana 2.1, a new Gemini 3 image generation and editing model built on Gemini 3.6 Flash that outperforms its predecessor Nano Banana Pro in character consistency, scoring 1106 +/- 14 on Multi-Character Consistency versus 1011 +/- 10 for Nano Banana Pro, with a 1M token context window, 4K token image output, and availability across the Gemini App, Google AI Studio, and the Gemini API, though it still struggles with blurry small-text rendering and imperfect character consistency.

• Google DeepMind publishes the model card for Nano Banana 2.1 on 6 October 2026, a Gemini 3 series image generation and editing model built on Gemini 3.6 Flash.
• Nano Banana 2.1 accepts text and image inputs with a 1M token context window and outputs 4K token images and 64K token text, and it is distributed through the Gemini App, Google AI Studio, the Gemini API, Google Search AI Mode, Google Ads, Google Flow, and Google Stitch.
• October 2026 evaluations show Nano Banana 2.1 (Thinking) scoring 1106 +/- 14 on Multi-Character Consistency versus 978 +/- 10 for Gemini 3.1 Flash Image ("Nano Banana 2") and 1011 +/- 10 for Gemini 3 Pro Image ("Nano Banana Pro"), while known limitations include blurry small-text rendering, imperfect character consistency, and rare persistent subject pose during editing.