nano-banana is a Large Vision Model (LVM) designed for fast and versatile image generation and editing tasks. It excels in maintaining character and background consistency across multiple variations, handling complex multi-step instructions for both image creation and editing, and delivering high-resolution outputs with fine detail retention.

Some other noteworthy features of nano-banana include text-to-image generation, local image editing (such as background or object removal and replacement), style transfer, and AI-powered upscaling and enhancement.

Metric	Value
Parameter Count	Unknown
Mixture of Experts	Unknown
Context Length	Unknown
Multilingual	Unknown
Quantized*	Yes

*Quantization is specific to the inference provider and the model may be offered with different quantization levels by other providers.