Google: Nano Banana 2.1

Google: Nano Banana 2.1 is an image generation and editing model from Google, built on the Flash tier. It succeeds Nano Banana 2 and Nano Banana Pro, and Google describes it as improving product recontextualization. It sits in both the image and text categories, so it is a multimodal model rather than a text-only or image-only system.

The model accepts image and text as inputs and can return image and text as outputs. Its context window is 65,536 tokens, with a maximum output of 58,982 tokens. Image generation is configurable: resolution can be set to 1K, 2K or 4K, and aspect ratio supports 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9 and 21:9. The n parameter accepts a single value, from 1 to 1, so each request produces one image. You can supply up to 14 input references, which is what makes editing and reference-driven generation possible.

Those capabilities point to a few practical patterns. Because the model takes image input and returns image output, it fits editing workflows where an existing image is adjusted rather than created from scratch. The reference limit of 14 supports tasks that need several source images at once, such as combining or restyling material, and the product recontextualization improvement suggests placing a product into new scenes or settings. Text output alongside image output means a single call can return generated visuals plus accompanying text. The resolution and aspect ratio options let you target different output shapes, from square and portrait to wide formats like 21:9, without post-processing crops.

You call the model through api14's OpenAI-compatible API using the model id "google/gemini-nano-banana-2.1". The supported request parameters are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, seed, structured_outputs, temperature, tool_choice, tools and top_p. Image-specific settings such as resolution, aspect_ratio, n and input_references are configured through the request as well. Because the interface is OpenAI-compatible, existing OpenAI client libraries and request shapes work with only the model id and parameters changed.

A few constraints are worth noting when choosing this model. The context window is 65,536 tokens, and n is fixed at 1, so requests that need multiple images per call are not supported here. If your task is purely text generation with no image input or output, a text-focused model may be a simpler fit. If you need more than 14 input references, this model will not accommodate that.