Google Nano Banana 2.1 release

Google released Nano Banana 2.1, a stable image generation and conversational editing model based on Gemini 3.6 Flash, on October 6. Its Gemini API documentation lists 1K, 2K and 4K outputs, multiple reference images and configurable thinking levels. Google says it improves realism, text rendering and character consistency compared with Nano Banana 2, but its model card also warns of blurry small text, imperfect consistency and occasional instruction-following errors. The Decoder reported a broader product rollout and compared the new API price with its predecessor. Performance comparisons remain Google’s evaluations or individual hands-on observations rather than a guarantee for every prompt.