Alphabet has recently introduced Nano Banana 2 Lite, a high-speed, low-cost AI model for image generation and editing aimed at developers. Officially named Gemini 3.1 Flash-Lite Image, it is now fully available for use.
This iteration emphasizes speed and cost-efficiency, capable of producing a standard 1K resolution image in approximately 4 seconds, with an average API call cost of just $0.034 per thousand images. In contrast, the standard Nano Banana model requires about 20 seconds for a similar image, and the Nano Banana Pro version's output pricing is roughly eight times higher. The company states this model is specifically designed for high-throughput commercial applications, rapid prototyping, and automated asset generation workflows.
Regarding capabilities, Nano Banana 2 Lite can swiftly generate scene sketches, preliminary data visualizations, and location-specific simulations based on world knowledge. It also features character consistency maintenance, making it suitable for applications like storyboard creation and e-commerce virtual try-ons. Furthermore, the model can render clear text directly within generated images, aiding teams in quickly validating multilingual advertisement layouts.
However, the "Lite" version does involve some trade-offs in features and image quality. It is limited to 1K resolution output and does not offer the 2K/4K multi-resolution options available in the standard and Pro versions. There are also noted limitations in rendering extremely small text and ensuring data accuracy in infographics. Despite this, evaluation data indicates Nano Banana 2 Lite achieved an Elo score of 1251 in text-to-image arena rankings, surpassing the first-generation Nano Banana's score of 1151 and even slightly exceeding the Nano Banana Pro's score of 1245.
The model is now accessible through Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform, allowing developers to begin using it immediately.
Comments