Google is changing access to Gemini, its artificial intelligence assistant, for personal accounts using the app and website. From October 9, 2026, free users will be limited to the Flash-Lite model. Access to the standard Flash and Pro models will require a paid subscription.

The changes also affect the entry-level AI Plus plan, but those subscribers will not necessarily switch on October 9. Google said it will email them with the date that applies to their account.

Which models each plan will offer

Under the announced structure, free accounts will lose both Flash and Pro. AI Plus subscribers will retain Flash-Lite and Flash, but no longer have access to Pro.

Google said AI Pro and AI Ultra subscribers will retain all three models. The Pro model will therefore be available through those two plans, rather than through the free or AI Plus tiers.

  • Free accounts: Flash-Lite only.
  • AI Plus: Flash-Lite and Flash.
  • AI Pro and AI Ultra: Flash-Lite, Flash and Pro.

The changes concern personal accounts. Work and school accounts are outside their immediate scope, and no timetable has been announced for similar restrictions on those accounts.

Effort settings and Deep Think

Google said users will be able to choose Low, Medium or High effort for each available model. These settings control how much effort the model puts into completing a task and developing its answer.

Higher effort is intended to produce more detailed responses, but it also consumes more of the user’s allowance. The exact date when the settings will become available to all relevant users has not been announced.

AI Pro and AI Ultra subscribers will have access to Deep Think, a mode for advanced parallel reasoning. Google lists that option alongside the three effort settings available for models.

How usage limits work

The access changes follow earlier adjustments to Gemini’s usage limits. Those limits are based on computing resources, with usage assessed according to the complexity of a prompt, the features used and the length of a chat.

Allowances refresh every five hours until a weekly cap is reached. Choosing a higher effort level uses more of that allowance, so the setting affects both response detail and quota consumption.