Gemini Nano Banana 2.1 Release and Resolution Upgrades
Google released Gemini Nano Banana 2.1 on October 6, 2026, introducing a high-efficiency image generation and conversational editing model designed to match Flash-level speed and cost efficiency. The update serves as a more efficient version of Gemini 3 Pro Image, also known as Nano Banana Pro, while building upon the architecture of Nano Banana 2, which is also identified as Gemini 3.1 Flash Image. ai.google.dev reported that the new model delivers major improvements in visual quality, prompt adherence, multi-turn character consistency, and text rendering.
New model fixes tiling artifacts in high resolutions
The updated model supports 1K, 2K, and 4K output resolutions, with 1K serving as the default setting. According to ai.google.dev, visual quality and realism have been enhanced across all three resolution tiers. The release addresses specific technical defects by fixing tiling artifacts that previously appeared in 2K and 4K outputs when generating images in wide and panoramic aspect ratios, specifically 1:4, 4:1, 1:8, and 8:1.
Multi-image fusion supports character consistency and object fidelity
Gemini Nano Banana 2.1 features multi-image fusion capable of handling up to 14 reference images simultaneously. ai.google.dev noted that this capability supports character consistency for up to four distinct characters and object fidelity for up to ten separate objects. The model also incorporates grounding through Google web and image searches to improve text rendering and infograph layout accuracy.
Configurable Thought Levels and Licensing
Users can adjust model behavior using configurable thought levels set to minimal, medium, or high, with medium serving as the default configuration. ai.google.dev stated that page content for the release operates under a Creative Commons Attribution 4.0 license, while code samples follow the Apache 2.0 license framework.
What are the Gemini Nano Banana 2.1 specifications?
What is the default output resolution for Gemini Nano Banana 2.1?
The default output resolution is 1K, though the model also supports 2K and 4K output resolutions with visual quality enhancements across all tiers, as reported by ai.google.dev.
How many reference images does the multi-image fusion feature support?
The multi-image fusion feature supports up to 14 reference images, allowing character consistency for up to four characters and object fidelity for up to ten objects, according to ai.google.dev.
What are the available configurable thought levels in the new model?
The configurable thought levels are minimal, medium, and high, with medium established as the default setting per ai.google.dev documentation.
Related reading