Today, Google introduced an updated version of its popular imaging model, the Nano Banana 2. The new model, which is technically part of the Gemini 3.1 Flash Image, is capable of creating significantly more realistic images than its predecessor.
In addition to the quality improvements, the Nano Banana 2 will now be the default primary model in the Gemini app for Fast, Thinking, and Pro modes.
Today, Google introduced an updated version of its popular imaging model, the Nano Banana 2. The new model, which is technically part of the Gemini 3.1 Flash Image, is capable of creating significantly more realistic images than its predecessor.
In addition to the quality improvements, the Nano Banana 2 will now be the default primary model in the Gemini app for Fast, Thinking, and Pro modes.
The new Nano Banana 2 retains the key high-resolution features of the Pro model, but generates images much faster. The company says that users can now create content with resolutions from 512px to 4K in a variety of aspect ratios.
Nano Banana 2 is capable of maintaining the consistency of up to five characters at once and maintaining the accuracy of up to 14 individual objects within a single workflow, opening up new possibilities for creating consistent stories.
With the launch of Nano Banana 2, it becomes the default image generation model in all Gemini tools. The company is also introducing it as the primary model for visual content creation in its Flow video editing tool.
In the Search segment, Nano Banana 2 will now be responsible for generating results through Google Lens and in AI Mode. This update covers 141 countries and is available in both the Google app and the web version for desktop and mobile.
For subscribers to premium plans — Google AI Pro and Ultra — there is still the option to use Nano Banana Pro for specific professional tasks. You can switch to it via the three-dot menu, selecting the image regeneration function.
Nano Banana 2 is available in preview format for the developer community via:
Gemini API and Gemini CLI;
Vertex API;
AI Studio platform;
The Antigravity development tool, which the company released last November.
The company said that all images created using the new model will be watermarked with SynthID, Google’s trademark for identifying AI-generated content, and will be compliant with the C2PA Content Credentials standard, an initiative developed by an industry group that includes Adobe, Microsoft, Google, OpenAI, and Meta.
