Google has added another image tool to its Gemini ecosystem with the launch of Nano Banana 2.1. The model is intended for both creating images and changing existing ones. Google says it has improved its ability to interpret instructions, produce detailed visuals and carry out edits without losing important elements. The company has also reduced the cost of image generation when developers use the model through its API.
At its core, Nano Banana 2.1 uses Gemini 3.6 Flash. Google is making the model available in more than one place rather than limiting it to a single application. Users can access it through the Gemini app, Google AI Studio and the Gemini API. It is also being added to AI Mode in Google Search, along with Google Ads, Flow and Stitch.
What Makes Google Nano Banana 2.1 Different?
The model is designed to handle prompts that contain a large amount of detail while keeping the requested elements under control. Google says this also applies when an existing image is being modified. Users can give the system several pictures to work from, after which it can take elements from those references and use them together in a newly generated image.
Nano Banana 2.1 accepts as many as 14 reference images during a single workflow. It can also remember the appearance of up to four characters and 10 separate objects. In practical terms, this means creators can reuse the same people, products or other visual subjects across a series of images without having them change noticeably from one image to another.
Nano Banana 2.1 Image Editing and Text Generation Features
Nano Banana 2.1 can change an existing picture from a simple text instruction. A user can alter a person’s clothes, replace the setting, remove unwanted objects or introduce new ones. The model can also give the picture a different visual treatment.
Google has worked on another old problem: words inside generated pictures. Nano Banana 2.1 is meant to produce clearer and more reliable text. That matters when someone needs a poster, diagram, infographic or another design in which the writing must remain readable. The system can generate images in 1K, 2K and 4K resolutions.
The model can also draw on Google Search and Google Image Search. This Search grounding feature gives it access to information from those services while it creates an image. The aim is to make the resulting picture more closely connected to real-world information.
Google Nano Banana 2.1 API Price and Cost Reduction
The price has fallen sharply. Google charges about $0.0336 for a 1K image through the API. The earlier model cost about $0.067. A 4K image now costs roughly $0.0756, compared with $0.151 before.
In both cases, the new price is close to half the old one. That difference matters most when a developer or business has to create thousands of images. At that scale, image generation can become a considerable expense. Lower API costs could therefore make such workloads easier to justify.
Google is now bringing Nano Banana 2.1 to its wider ecosystem. Developers and ordinary users will gain access to its newer tools for creating and changing images as the rollout continues.
Final Words
For those who spend too much time trying to get AI-generated images to behave, Google’s Nano Banana 2.1 appears to be a pretty serious upgrade. The model, developed on the Gemini 3.6 Flash platform, will provide enhanced instruction-following abilities, more detailed visuals and powerful image editing, and maintain consistency of characters and objects across multiple creations.
It could be especially helpful for posters, infographics, product images, and other creative projects, due to its capacity to handle up to 14 reference images and produce clearer text.The more surprising part of the news, though, is the price. Nano Banana 2.1 may be much more budget-friendly for businesses that are creating images on a large scale, as the API costs have dropped by about half.
With Google’s rollout of access to Gemini, AI Studio, Search, and more, Nano Banana 2.1 might soon be one of the most useful image generation tools for Google.
