Google details October Gemini access changes and compute-based usage limits

Google has updated its Gemini Apps access guidance with changes to model availability for personal accounts starting in October. The company says the changes begin on October 9 for people without an AI subscription. AI Plus subscribers should receive an email explaining when the changes take effect for their account.

The guidance also explains how Gemini's compute-based usage limits work. That is important context for anyone comparing plans: the amount of work a request requires matters, not just the number of messages sent. A short exchange and a demanding task using premium features need not consume the same share of a user's allowance.

Check account timing before changing a workflow

Google distinguishes free-account rollout timing from the instructions sent to AI Plus subscribers. The October 9 date therefore should not be treated as a single universal switch for every Gemini user. The notice concerns personal accounts; it does not establish that an enterprise account or developer API has the same access rules.

Google advises using the latest version of the Gemini mobile app. For someone relying on the app for a recurring task, the practical first step is to check the available model choices in the current app and read the notice for the account's plan. A name shown in an announcement does not by itself establish that a particular account has access.

Usage is tied to computation

The help page says usage calculations take account of prompt complexity, the features used and the length of the chat. It describes limits that refresh every five hours until the weekly limit is reached. Paid users have higher allowances than people without a Google AI subscription.

A five-hour refresh is not the same as unlimited access every five hours. The weekly limit remains part of the described system. Readers planning repeated demanding work should consider both the shorter refresh window and the overall allowance, rather than budgeting around one of them alone.

The notice identifies media generation, Deep Research, the Pro model, extended thinking and Deep Think as examples of functions that require more usage. This explains why counting messages can be a poor way to estimate how long a plan will support a workflow. Two users sending similar numbers of prompts may be asking Gemini to perform very different amounts of work.

Reasoning effort changes the tradeoff

Google says users can select low, medium or high effort for each available model. Higher effort can improve the model's ability to complete tasks and provide more thorough answers, while also using more of the allowance. AI Ultra subscribers additionally have access to Deep Think for maximum parallel reasoning.

That gives users another decision beyond selecting a model. For a straightforward request, a higher-effort setting may consume capacity that would be more useful on a difficult task later. For a demanding request, choosing a lower setting merely to save allowance may not produce the result the user needs. The useful comparison is the completed task at the chosen setting, not the setting name alone.

Compare plans against the work you actually do

The published usage table describes AI Plus as twice the standard allowance and AI Pro as four times the standard allowance. AI Ultra is listed as five or twenty times AI Pro depending on the subscription. Those are relative plan allowances in Google's guidance, not a fixed guarantee of a particular message count.

The notice also links plan-management options for subscribing, upgrading, changing or cancelling a plan. Before paying for an upgrade, check the account's current access and the features the workflow needs. Google's revised guidance makes the connection between model access, effort and usage clearer, but the account's actual available options remain the most useful starting point.