Gemini Notebook is replacing daily limits with a compute-based system on 2 September

The announcement is short: a single post from Yesul Shin, Product Manager for Gemini Notebook, framed entirely around user choice. "This change will give you more choice over how you want to spend your limits and manage your tasks," the post says, and points to a genuinely useful addition alongside it, a "Generate later" option that defers a resource-heavy output (Video Overviews, Slide Decks) until your quota refreshes, with a notification when it's ready. Google's own Gemini Notebook Help page has already been rewritten to describe the new system, three days ahead of the rollout, which is where the weekly-ceiling detail comes from that the blog post itself doesn't mention.

What's actually changing

The current system, still live as of this writing, counts a flat number of actions per calendar-ish day: 50 chat queries and 3 Audio Overviews on the free Standard tier, up to 5,000 chats and 200 Audio/Video Overviews a day on Ultra 30TB. Every action costs the same regardless of how demanding it actually was, a short one-line question and a 40-source cross-document synthesis both counted as "one chat." Our full limits table breaks down every tier's current numbers if you want the before picture while it's still accurate.

The new system replaces that flat count with a shared budget, described identically across Google's blog post and support page as factoring in "prompt complexity, the models and features you use, the length of your chat and specific features." A quick one-line chat question and a Video Overview generated from 40 sources will not cost the same amount against your quota anymore, the video will draw far more. This is the same underlying shift the Gemini app went through on 17 May 2026, when Google eliminated its own flat daily prompt counts in favor of a rolling compute quota, covered in our Gemini Deep Research comparison.

• No plan (free): standard baseline limits, Google's terminology for the lowest tier.
• Google AI Plus: 2x the standard baseline.
• Google AI Pro: 4x the standard baseline.
• Google AI Ultra: 5x or 20x on top of Pro, not on top of standard, depending on whether you're on the $99.99/mo 20TB plan or the $199.99/mo 30TB plan. That distinction is easy to misread at a glance, since Google's support page lists it right after the Pro line without repeating "relative to Pro" on the Ultra line itself.

How the 5-hour window replaces the daily reset

Today, your daily allowance resets "after 24 hours," language Google has never tied to a specific clock time or timezone, which is itself a long-running source of confusion covered in our limits guide. The new system resets your compute budget every 5 hours instead, up to a weekly ceiling, per Google's support page. In practice that means a heavy research session early in the day doesn't necessarily lock you out until tomorrow; you may get a fresh partial budget again within hours, though Google doesn't say the weekly ceiling itself resets on a predictable schedule either.

"Generate later" is the other real addition. If an output would put you over budget, Gemini Notebook now offers to queue it instead of just refusing outright, generating automatically once you're back under your limit and notifying you if you have notifications turned on. Google's support page notes this is web-only for now, with no stated timeline for the mobile apps.

What Google still hasn't told you

Every source covering this, the blog post, Google's own support page, and independent reporting from 9to5Google and Android Police, repeats the same multipliers and the same absence: no actual number. Nobody, including Google, has published what a "standard" 5-hour or weekly budget translates to in real terms, how many chats, how many Audio Overviews, before you hit it. Any article telling you the exact new free-tier limit as of today is guessing. Android Police's own headline calls the replacement system simply "more complicated," a fair read given that a fixed number you could count in your head is being swapped for a budget you can't see until you've spent some of it.

Also unconfirmed: whether the reset-timing ambiguity that already dogs the current system, is the daily reset really tied to your first use of the day, or a fixed clock time nobody publishes, carries over into the new 5-hour window, or whether Google is finally giving Gemini Notebook the kind of visible usage dashboard it built for the Gemini app after May's rollout (more on that below). Neither the blog post nor the support page addresses it.

Read the fine print: this already happened once, in May

The detail missing from most coverage of this announcement is that Google already ran this exact experiment, on the Gemini app itself, on 17 May 2026, and had to walk part of it back. Within days, 9to5Google reported Google was adjusting the new system in response to complaints: capping how much of an individual prompt's quota usage could count against the total, making the lightweight Gemini 3.1 Flash-Lite model exempt from the compute quota entirely, doubling the Omni generation allowance for Ultra subscribers, and fixing a bug where one or two Omni video generations could drain a user's entire quota in one shot. Google's own statement on failed requests at the time: "If a request fails, you won't be charged. Our system mistakes are on us, not you."

The complaints weren't abstract. One Google AI Pro subscriber reported a single avatar-video prompt burning an entire 5-hour allowance and still failing, prompting Gemini's own product lead, Josh Woodward, to reply publicly: "Yikes, let us take a look!" The underlying frustration, as one contemporary write-up put it, was that paying subscribers now had to "think like cloud-cost managers" instead of just using the product, since there was no way to know a prompt's cost before running it.

What to actually do before 2 September

• Don't panic-generate everything now. The current flat daily counts stay in effect until the rollout date; there's no reason to burn your Audio Overview cap early trying to "beat" the change.
• Expect Studio outputs to cost more than chat. Since the new budget explicitly weighs "features you use," a Video Overview or Slide Deck built from a large source set is a reasonable bet to eat more of your quota than a short chat question, mirroring how the Gemini app's Omni video generation turned out to be the expensive action there.
• Use "Generate later" instead of retrying. If you hit your limit on web, queue the output rather than repeatedly retrying, which likely won't help and may not be free of cost either.
• Export anything you'll need before quota gets tight. If a compute-heavy output like a Report or Audio Overview is going to cost more under the new system, save your current work now rather than after 2 September; our NotebookLM to PDF guide covers exporting notes and outputs before you're rationing generations.
• Watch for a follow-up adjustment. Based on the Gemini app precedent, treat the first one to two weeks after 2 September as the period most likely to see a correction if the initial budget proves too tight for common workflows.
• Check your own tier's real multiplier, not just the headline number: Ultra's 5x/20x is measured against Pro, not against the free baseline, which changes the effective gap between Pro and Ultra more than a naive reading suggests.

People also ask

When do Gemini Notebook's new usage limits start?

September 2, 2026, for consumer accounts on web and mobile, per Google's own blog post and support page.

What is Gemini Notebook's new usage limit, exactly?

Google has not published an exact number for any tier. It has published multipliers relative to the free "standard" baseline: AI Plus is 2x, AI Pro is 4x, and AI Ultra is 5x or 20x on top of Pro (not on top of standard), depending on the Ultra plan.

Does the daily limit reset go away entirely?

Yes. The new system refreshes your compute budget every 5 hours instead of once every 24, up to a weekly ceiling that Google also hasn't put an exact number on.

What is "Generate later" in Gemini Notebook?

A new option that lets you queue a resource-heavy output, like a Video Overview or Slide Deck, when you're over your limit. It generates automatically once your quota refreshes and can notify you when it's ready. It's web-only as of the announcement, with no confirmed mobile timeline.

Did this happen to the regular Gemini app too?

Yes, first, on 17 May 2026. That rollout drew enough complaints about unpredictable costs that Google adjusted it within about ten days, exempting a lightweight model from the quota, capping per-prompt usage counted against the total, and fixing a bug where a single video generation could drain an entire quota window.