Google is changing how Gemini Notebook usage limits work from September 2, dropping the fixed daily count of feature generations in favour of a compute budget that refreshes every five hours. The practical effect is that how much you can do in a day now depends on how demanding your work is, not on a number you can look up in advance.
Announcing the change on Google’s blog, Gemini Notebook product manager Yesul Shin framed the five-hour reset as the headline benefit.
“And since limits refresh every five hours rather than daily, you can now keep working throughout the day.”
What compute-based Gemini Notebook usage limits actually measure
Google says your overall limit will factor in prompt complexity, chat length, number of sources, and the features you use. That is four separate variables, and none of them is something you can count while you work. A short question against two sources costs less than a long back-and-forth against twenty; a Video Overview costs more than a chat reply. Under the old system you knew you had, say, a set number of generations per day for your tier. Under this one you find out where the ceiling is by hitting it.
Google’s mitigation for that is visibility rather than predictability. Notebooks will show a usage status, and a fill bar appears under Studio generations to indicate expected cost — the more it is filled, the more the output is expected to cost you. If a request would push you over, Notebook is meant to suggest a cheaper alternative output instead.
There is also a deferral mechanism. If you run out mid-task, heavy outputs like Video Overviews and Slide Decks can be queued with “Generate later”; they build automatically once your allowance resets, and you can set a notification for when they are ready. 9to5Google reports that “Generate later” arrives on the web first.
How the limits scale by subscription
Google’s own post does not publish per-tier numbers. The multipliers below come from 9to5Google’s reporting, and they are relative — Google has not stated what a “standard” limit is in absolute terms, so there is no baseline to multiply.
- Free / no plan — standard limits
- Google AI Plus — 2x standard
- Google AI Pro — 4x standard
- Google AI Ultra — 5x or 20x higher than AI Pro, depending on the subscription
That structure is worth reading for what it does rather than what it is called. “Flexible” is Google’s word for a limit that moves, and a limit that moves is one a free user cannot plan around and a paying user cannot easily compare between tiers. It is also the same compute-metering model Google moved the main Gemini app to earlier this year, which suggests the direction of travel across the whole product line rather than a one-off tweak to Notebook.
When it arrives
The change rolls out to consumer accounts on web and mobile starting September 2. Google has not tied it to a version number or a staged regional rollout in its announcement. All existing features stay available; what changes is the accounting behind them. Notebook has been picking up heavier outputs recently — it can now pull in Google Play Books ebooks you own as sources — and those are exactly the workloads a compute budget prices highest.
Source: 9to5Google (tier multipliers)
