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

Google is rolling out a substantial structural update to the Gemini app experience, introducing significant changes for users on the free tier alongside a broader rollout of new thinking and reasoning controls for subscribers. The updates, which began appearing on the vast majority of free accounts late in the evening Pacific Time, represent a major shift in how Google manages its artificial intelligence models across different user tiers.

Under the new system, free users—referred to by Google as those "without a plan"—will no longer have the ability to manually switch between different Gemini models. Previously, users could choose specific models from a picker menu depending on their preference or task complexity. Moving forward, the official model picker option for free users is strictly locked to "Auto," utilizing the Gemini 3 model family. According to Google’s support documentation updated last week, this automated approach relies primarily on the (3.5) Flash-Lite model.

Google explains that the auto model selection feature is designed to dynamically route each incoming prompt to the specific Gemini model that suits it best. Under this automated framework, the vast majority of standard prompts will be directed to Flash-Lite, which stands out as Google’s fastest model. However, for queries that demand deeper analytical processing or more complex reasoning, the system can automatically scale up, routing prompts to either the standard Flash or Pro models as required.

For users who prefer a more predictable experience, Google has clarified how the system behaves when the automated selection feature is adjusted. Official documentation states that if smart model selection is turned off, chats for users without a paid AI plan will default exclusively to Flash-Lite for all responses, ensuring rapid generation times at the expense of advanced reasoning capabilities.

Concurrently, Google is introducing a revamped system of "effort levels" to replace the older "Extended" thinking option within the application interface. These newly structured choices closely align with the established "thinking levels" found in Google AI Studio and Antigravity, giving users finer control over how hard the AI works to solve a problem. While these effort levels are rolling out broadly to support all subscribers, they arrive alongside tier adjustments that will soon restrict Google AI Plus users to specific model options, namely Flash-Lite and (3.8) Flash.

The speed of the deployment has surprised many users, with the changes to free accounts appearing rapidly across supported devices and platforms. Google’s rapid rollout strategy indicates a concerted push to standardize the Gemini application experience around its newer, more efficient model architecture, emphasizing automated routing over manual user configuration for the casual tier. As these updates continue to propagate across global servers, users on both free and paid plans will notice a distinct evolution in how they interact with Google’s artificial intelligence ecosystem.