Google Overhauls Gemini App for Free Users with Automatic Model Selection and New Effort Levels

Google is rolling out a sweeping update to the Gemini application experience, introducing significant changes for users on the free tier alongside new controls for everyone. The update, which follows announcements made by the company regarding model limits, shifts how free users interact with Google’s artificial intelligence ecosystem by automating model selection and phasing out manual picking in favor of a streamlined approach.

Under the newly deployed system, free users who do not subscribe to an AI plan no longer have the ability to manually switch between different Gemini models. Instead, their model picker defaults exclusively to an "Auto" setting powered by Gemini 3 models, which Google’s support documentation identifies as the (3.5) Flash-Lite model. This automatic model selection dynamically routes user prompts to the specific Gemini model deemed best suited for the task at hand.

According to Google, the majority of standard prompts will be automatically directed to Flash-Lite, which serves as the company’s fastest model. However, for queries that require deeper reasoning, complex analysis, or advanced problem-solving capabilities, the system can dynamically route prompts to more powerful models such as Flash or Pro. The underlying philosophy of this change is to abstract the technical complexity of model management away from everyday users, ensuring that they receive optimal responses without needing to understand the underlying infrastructure.

For users who prefer a uniform experience, Google has built in a specific behavior regarding smart model selection. Official documentation notes that if smart model selection is turned off, chats for users without an active AI plan will consistently rely on Flash-Lite for all generated responses, maintaining a predictable speed and output baseline across sessions.

Alongside the structural changes to the free tier, Google is introducing a new suite of "effort levels" across the platform. These new controls are designed to replace the previous "Extended" option that users previously relied on for more thorough generations. The three distinct choices align closely with the "thinking levels" found in Google AI Studio and Antigravity, giving users greater nuance over how much computational effort the model should dedicate to processing a prompt and formulating a response.

The rollout of these features is currently underway. Observations from various accounts indicate that the new free tier experience is appearing across the vast majority of free accounts, reflecting a rapid deployment schedule by Google. Meanwhile, the new effort levels are also making their way to all subscribers as the backend updates propagate globally.

In tandem with these adjustments for free and standard users, ecosystem changes are touching paid tiers as well. Google AI Plus users will soon see their model access streamlined, with upcoming limitations restricting their primary experiences to Flash-Lite and (3.8) Flash. As these updates continue to deploy, Google aims to create a more integrated, automated, and responsive AI assistant experience that scales intelligently depending on user subscription status and task complexity.