Field note
Gemini Usage Limits: Five-Hour and Weekly Resets
Gemini app limits are compute based, refresh every five hours and cap out weekly, so a small team should plan model choice and fallbacks rather than expect a fixed number of prompts.
Gemini app limits are compute based, refresh every five hours and cap out weekly, so a small team should plan model choice and fallbacks rather than expect a fixed number of prompts.
How the limit works
Google's help page for Gemini Apps limits describes them as compute-based rather than a count of prompts. At the time of writing, the limit takes into account the complexity of the prompt, the models and features used, and the length of the chat. The limit refreshes every five hours until the weekly limit is reached. Google flagged a change to usage limits from 17 May 2026, so check the page rather than relying on older screenshots.
Paid plans raise the ceiling. Google states that Google AI Plus gives two times the standard limits, Google AI Pro four times, and Google AI Ultra five or twenty times the Pro limits depending on subscription. Subscribers who reach their limit can continue the conversation with Flash-Lite. Google also says that more advanced models and higher thinking levels consume more of the allowance, as do features such as media generation and Deep Research.
Two caveats are written into the same page. Limits may change without notice, including because of capacity, and when demand is high users without a paid plan may be limited before subscribers. Some features such as Deep Research may be unavailable to free users during busy periods.
That page covers personal accounts. Work accounts are governed by the Workspace edition, and the Workspace admin help describes an AI Expanded Access add-on for users who need more capacity, along with a page on AI usage limits for administrators.
How to see usage
Gemini notifies the user when they are close to a limit and, once reached, tells them when it refreshes. The current position is under Settings, then Usage Limits, in the Gemini app. There is no documented per-user dashboard for personal accounts beyond that; for Workspace, the admin help includes a page on reviewing Gemini usage in the organisation.
A simple habit works for a small team: when the limit message appears, note the time. The five-hour window refreshes from there. The weekly limit does not, so a person who hits it on Tuesday is on Flash-Lite, or on nothing for free accounts, for the rest of the week.
Why a team hits the limit faster than expected
- One long chat all day. Chat length is a factor, so a single conversation carried across many tasks costs more per prompt than a fresh chat per task.
- Pro with Extended thinking for trivial jobs. Google's own guidance is that Standard thinking is best for most questions. Pro and higher thinking levels cost more allowance.
- Large uploads. The context window is 32k tokens for Flash-Lite, 128k for Flash and 1 million for Pro at the time of writing; reading a large file each turn is expensive.
- Deep Research and image generation consume more than chat.
- Shared logins. One account used by five people burns one person's allowance and breaks the terms and any audit trail.
Planning for a small team
| Need | Plan |
|---|---|
| Daily drafting and summaries | Flash, new chat per task, Standard thinking |
| Occasional complex analysis | Pro, scheduled rather than ad hoc, one chat per piece of work |
| Deep Research | Batched, a small number per day, results handed to a person to verify |
| Deadline-critical work | A written fallback: the Workspace side panel, another assistant, or doing it by hand |
Give each person their own account. Track who hits limits and when for a month. If one person hits the weekly limit regularly, that is evidence for a plan upgrade for that person, not a reason to share a login. Upgrades for personal accounts are through a Google AI plan; for work accounts they are through the Workspace edition or the add-on.
The Gemini API is a separate system with separate limits: requests per minute, tokens per minute and requests per day, applied per project and rising by usage tier, with a 429 response when exceeded. Current values are shown in AI Studio. Do not mix API planning with app planning.
The continuity pattern for any assistant is in AI chatbot usage limits and business continuity, and the capacity worksheet in ChatGPT rate limits: team capacity plan transfers directly.
Separating a limit from a fault
A limit message is explicit and includes the refresh time. Anything else, such as a generic error, a blank response or a spinner, is a fault or a capacity event, not a limit. Check the Google Workspace status dashboard, then follow a proper isolation sequence before assuming the allowance is gone. Buying a higher plan does not fix a network block.