What is Nano Banana 2.1 and how does it differ from the previous version?
Google has introduced a new model for image generation and editing, Nano Banana 2.1. The model runs on Gemini 3.6 Flash and replaces the February Nano Banana 2, offering better visual content quality at significantly lower costs.
The developer description states that the update increases text reproduction accuracy and improves character consistency. Google also bills the release as a more efficient analogue of the Pro version while acknowledging existing limitations.
What are the model’s technical capabilities and limitations?
Nano Banana 2.1 supports a context window of up to 1 million tokens on input. The model can process up to 14 reference images simultaneously and provide visual consistency for four characters and ten objects within a single session or dialogue.
Alongside its capabilities, the developers note functional limits. The model can sometimes misplace objects spatially, and fine text in 1K images may remain blurry. Additional technical details are available in the model card.
How have image-generation rates changed and where is the model available?
The price to generate a 1K image has dropped by almost half, from $0.0670 to $0.0336. For comparison, a 4K image now costs $0.0756 instead of $0.1510. The flagship Nano Banana Pro remains more expensive at $0.1340 per 1K image.
The model is already available via the Gemini API, the Google AI Studio Gemini app and other ecosystem tools. In addition, the tool can connect to Google Search and lets users choose one of three levels of reasoning depth, which affects the detail of the final image.
