Quick Answer
Gemini Apps has a short window and a weekly ceiling. Usage refreshes every five hours only until the weekly limit is reached. Media generation, Deep Research, advanced models, and higher thinking levels can consume the allowance faster.
After a paid user reaches a limit, Flash-Lite can remain available for conversation. A Plus user who needs paid-model, video, or Deep Research access again must wait for the applicable reset or move to a higher plan; Plus cannot purchase AI credits.
The word that matters is “until”
Google’s Gemini Apps limits guide says limits account for prompt complexity, the model and features used, and chat length. It then states that the limit refreshes every five hours until the weekly limit is reached.
That creates two different states:
- Five-hour wall, weekly allowance remains: wait for the displayed short-window reset or upgrade.
- Weekly wall reached: another five-hour interval does not restore the exhausted weekly allowance; wait for the weekly reset shown in the app or upgrade.
Google does not publish a fixed prompt count or token allowance for these tiers. It also warns that limits can change without notice, including because of capacity constraints. The in-app meter and reset time are the operational source.
The same five-hour sentence appears in the personal-account comparison discussed by the personal Gmail and Education terms guide. That guide answers which account terms apply. This one answers how to react to the two quota walls.
Read the multipliers against the correct baseline
The current limits page describes plan access as:
| Plan | Published relative limit |
|---|---|
| No Google AI subscription | Standard limit |
| Google AI Plus | 2× the standard limit |
| Google AI Pro | 4× the standard limit |
| Google AI Ultra | 5× or 20× the AI Pro limit, depending on the Ultra subscription |
Ultra’s multiplier is relative to Pro, not the free standard limit. Do not flatten the table into one shared baseline.
These are relative allowances, not a guarantee of a specific number of prompts. Google can adjust them with demand and capacity.
Heavy features push both meters faster
Google’s model-access change notice names media generation, Deep Research, Pro models, extended thinking, and Deep Think as examples that consume more usage. A five-hour window can therefore disappear much faster during video or research work than during light chat.
The official pages do not publish a universal percentage for one generated video or one report. Any third-party “four videos per window” measurement is a snapshot, not a rate card, so it is not used here.
Deep Research also has feature-specific limits. Google’s Deep Research guide says the app limits daily research requests and simultaneous requests, shows the remaining count near the limit, and gives Pro and Ultra users more reports. That feature counter can be reached separately from the five-hour and weekly compute limits.
What remains after the wall
Google says a subscriber who reaches a model limit can continue conversations with Flash-Lite. That is a fallback for ordinary chat, not a restoration of the paid model or heavy feature that reached the limit.
| Plan | Five-hour or weekly limit reached | AI-credit purchase |
|---|---|---|
| No subscription | Wait, use the available fallback, or subscribe | Purchase page does not list this tier |
| AI Plus | Use Flash-Lite where offered; wait for the applicable reset or upgrade | Not available |
| AI Pro | Use fallback, wait, upgrade, or buy credits only if all purchase gates are satisfied | Conditionally available |
| AI Ultra | Use fallback, wait, or buy credits only if all purchase gates are satisfied | Conditionally available |
Google’s AI-credit purchase page explicitly excludes Plus. Pro or Ultra is necessary but not sufficient: billing channel, country, family role, and other gates can still block the purchase.
Those gates and the cost of leaving Apple, partner, or Pixel Pass billing are separated into the Google AI credit channel-switch guide. Do not turn a quota-wall problem into a billing migration until the purchase matrix has been checked.
A two-meter operating routine
Before a heavy Gemini session:
- Open Gemini Apps Settings → Usage limits and record both displayed reset times.
- Route light chat away from the Pro model when it does not need advanced capability.
- Schedule video generation and Deep Research early enough that a weekly wall will not block delivery.
- If only the five-hour wall is full, use Flash-Lite or wait for that reset.
- If the weekly wall is full, stop expecting another five-hour refresh to restore the paid allowance.
- If on Plus, compare waiting with upgrading; do not plan around a credit purchase that Google excludes.
- If on Pro or Ultra, confirm every credit-purchase gate before promising an immediate top-up.
The practical distinction is simple: the five-hour bar can recover while weekly capacity remains. The weekly bar is the hard ceiling.
Sources
- Gemini Apps limits and upgrades — five-hour and weekly mechanism, plan multipliers, fallback, and capacity warning; checked 28 August 2026
- Changes to Gemini model access and limits — compute-based limits and heavy-feature examples; checked 28 August 2026
- Purchase Google AI credits — Plus exclusion and eligible-plan boundary; checked 28 August 2026
- Use Deep Research in Gemini Apps — daily and concurrent research limits; checked 28 August 2026
- Google AI subscriptions — current tier comparison; checked 28 August 2026