Model
Nano Banana 2.1.
Power
—
API cost
In plain English
Google’s current model for generating and editing images. You describe a picture or upload an existing one, then use ordinary language to create or change it.
Good for
Product images, visual concepts, adverts, image editing, diagrams, and social content. It is particularly useful when several rounds of changes need to keep the same subject or product.
Things to be aware of
It can still alter faces, products, positioning, or small details that you asked it to preserve. Small text and long paragraphs can come out badly. Check consent, copyright, brand rules, and whether the final image could mislead somebody.
What Google DeepMind says
“Creation and editing of images with professional levels of precision and control and multiple, quick iterations.”
Seen here: Google DeepMind’s Nano Banana 2.1 model card · Model Card — Google DeepMind
What everyone else says
Independent tests suggest that 2.1 is strong at editing and is cheaper than earlier versions. The improvement is not universal. In one test, reviewers could not reliably distinguish its photorealistic images from Nano Banana Pro. Nano Banana 2.1 review · TechRadar comparison · eesel AI
- Context window
- Up to 1.0M tokens
- It appears in products including Gemini, Google AI Studio, Flow, Ads, and Stitch. Access and limits depend on the product and subscription. Model Card — Google DeepMind
- Knows up to
- March 2026
- although some knowledge may stop at January 2025
- Run it yourself
- No, online services and API only