Google is changing consumer usage limits in Gemini Notebook from September 2, 2026. Instead of treating every prompt or generated artifact as an equal unit, the service will consume a shared compute allowance based on the work requested.
Prompt complexity, chat length, number of sources, model choice and output type can all affect usage. Limits refresh every five hours until the weekly allowance is reached, and the interface will show remaining usage and the relative cost of heavier Studio generations.
Google is also adding a Generate later option for expensive outputs such as Video Overviews and Slide Decks. Those jobs can wait for quota to become available and complete later.
The trade-off is predictability. Lightweight work may become more flexible, but users can no longer assume that two requests consume the same amount of allowance. Google has not published a universal compute-unit price table, so the in-product meter becomes the practical reference.