Nano Banana: Google's new AI image model

Google's Nano Banana image model generates in 1-2 seconds and takes edit instructions in plain language. It's available in Denmark through Gemini 2.5, and API access costs about 0.30 kr per image.

Nano Banana: Google's new AI image model

Google's newest AI image model is called Nano Banana. The name is silly. The model is not. It's available in Denmark now through Gemini 2.5, and it handles image generation and editing better than most of what came before it.

From mystery codename to public release

The model first appeared on LMArena, the anonymous benchmark platform, under the codename "nano-banana". It beat the competition there without revealing who built it, and beta testers around the world tried to guess who was behind it.

The name reportedly started as an internal nickname during development at Google DeepMind. It stuck.

How the model works

Nano Banana is built on a multimodal vision-language architecture. It combines image understanding with natural language processing, which means you edit images by describing the change you want.

"Make the sky more dramatic." "Move the chair a bit to the left." "Add a smile to the person's face." The model takes instructions like that and executes them with surprising precision.

It's also fast. A typical generation takes 1-2 seconds per image. Competing models often need 10-15 seconds or more.

What it can do

Four features stand out:

• Character consistency: the same person or figure stays recognizable across a series of images.
• Multi-image fusion: it can combine up to 13 separate image elements into one coherent photo.
• Iterative editing: you can make change after change without the model "forgetting" or degrading earlier edits.
• World knowledge: the model knows not just what things look like, but how they work in reality.

How it compares

Against OpenAI's DALL·E 3, Nano Banana consistently produces sharper, brighter images with better detail.

Midjourney is still the king of spectacular, artistic AI art. Nano Banana wins on practical use and editability.

Against Stable Diffusion, the difference is setup. Nano Banana delivers stable, consistent output without requiring technical expertise.

Where it's already in use

• E-commerce: companies generate product images automatically
• Architecture: firms use it for fast visualizations of their designs
• Teaching: teachers create custom illustrations for their lessons
• Game development: indie developers generate thousands of unique NPC portraits

Access and pricing in Denmark

The model is fully available in the Gemini app in Denmark, for both free and paying users. For developers, API access costs about 0.30 kr per generated image.

The absurd name covers a useful tool. If you produce images for anything, product photos, presentations, teaching material, it's worth an hour of testing. The free tier means that hour costs you nothing.