Google is making changes to user access to its Gemini AI models under which free users and lower-cost subscribers of the service will face limitations in choosing AI models. Conversely, subscribers to more expensive plans will receive more advanced features.
According to Itna citing 9to5Google, based on an updated Google support document, from October 9, 2026, users without a paid subscription will no longer have access to the Flash and Pro models and will only be able to use the Flash-Lite model. This model is designed to run tasks with lower computational resources.
Subscribers to the Google AI Plus plan, who pay $4.99 per month, will be limited to the Flash-Lite and Flash models and will lose access to the Pro model. Google said these users will be notified by email about the timing of the changes.
On the other hand, subscribers to Google AI Pro, at $19.99 per month, and users of the AI Ultra plan will still have access to all three models. In addition, the Deep Think capability, previously reserved for higher-tier plans, will also be made available to AI Pro subscribers. This capability is designed to solve complex problems using Parallel Reasoning.
Google also intends to introduce three levels of compute effort - low, medium, and high - for models available in the Gemini app in October. Choosing a higher level can produce more accurate and comprehensive responses, but it also increases usage quota consumption.
It is not yet clear what position the new Gemini 4 Argon model will occupy within this structure. Google has stated that this model will first be released to AI Ultra subscribers, but its final subscription level has not yet been determined.