Nano Banana 2.1 halves the price of an AI image. Read the fine print.
Google's new image model costs about half as much per picture, but input and text tokens now cost more. A zine-style explainer.
Google's new image model costs about half as much per picture, but input and text tokens now cost more. A zine-style explainer.
· Captions, from Notebookcheck, FourWeekMBA, XenoSpectrum
Collage of public-domain archive photos via Wikimedia Commons (see the assets branch credits), xerox and halftone treatment by AI Broadsheet; ink cats, eyes and banana drawn by AI Broadsheet. Music: “Tropicorp Advertisement” by code_box (CC0, Freesound); paper and stamp sounds: Freesound (CC0).
What the video says
Google's new image model, Nano Banana 2.1, makes pictures for about half the price. On the Gemini API, a 1K image drops from 6.7 cents to about 3.
4. A 2K image, from about 10 cents to 5. But read the fine print. Input now costs a dollar fifty per million tokens, three times as much.
Text and thinking output rose from three dollars to seven fifty. So if you send lots of reference images or long prompts, you'll save less than the headline suggests. Google also lists its limits. Small text often stays blurry at 1K, and the model sometimes mixes up left and right.
Independent tests aren't out yet. Cheaper pictures, pricier words.
Sources
Collage of public-domain archive photos via Wikimedia Commons (see the assets branch credits), xerox and halftone treatment by AI Broadsheet; ink cats, eyes and banana drawn by AI Broadsheet. Music: “Tropicorp Advertisement” by code_box (CC0, Freesound); paper and stamp sounds: Freesound (CC0). Report a mistake