Google Rolls Out Major Changes to Free Gemini App Experience
Google is officially rolling out major changes to the <a href="https://9to5google.com/guides/gemini-app/">Gemini app</a> experience for users on the free plan, bringing shifts in how models and thinking levels are handled.
Radwan AbdulhadiEditor and technology writer
Free Plan Model Limitations and Auto Selection
As <a href="https://9to5google.com/2026/10/03/gemini-model-limits-oct-26/">announced last week</a>, Google is actively rolling out a substantial structural shift to the <a href="https://9to5google.com/guides/gemini-app/">Gemini app</a> experience tailored specifically for users operating on the free tier. Under the updated framework, free users who do not have an active subscription plan can no longer manually choose or switch between different individual models within the interface.
Formally, the picker interface now leaves 'Auto' with Gemini 3 models as the sole option available to free tier individuals. Support documentation published by Google previously referred to this baseline capability as Flash-Lite, outlining a streamlined operational approach for everyday requests processed through the <a href="https://9to5google.com/guides/gemini/">Gemini</a> platform ecosystem.
How the Auto Model Selection Operates
Explaining the technical mechanics behind the new architecture, Google states that the system will leverage automatic model selection to route every individual user prompt directly to the underlying <a href="https://9to5google.com/guides/gemini/">Gemini</a> model that matches the specific contextual requirements of the request.
Under this automated routing strategy, the vast majority of standard prompts are directed to Flash-Lite, which functions as the platform's fastest model. However, prompts that require significantly deeper analytical reasoning may dynamically route to either Flash or Pro depending on the underlying complexity assessed by the system.
Furthermore, if a user decides to turn off the smart model selection feature entirely, all active chat sessions for accounts without a dedicated AI plan will systematically default to utilizing Flash-Lite for generating every single response.
Introduction of New Effort and Thinking Levels
Alongside the model routing changes, Google is also introducing fresh 'effort levels' across the application to completely replace the previous 'Extended' option previously familiar to regular users.
These three newly implemented choices are designed to closely align with the established thinking levels currently found within Google AI Studio and Antigravity environments, offering a standardized experience across developer and consumer tooling.
Observations from active deployments indicate that these new effort levels are currently rolling out to all subscribers alongside the free tier updates, with Google pushing out the package rapidly across accounts.
Rollout Speed and Subsequent Tier Restrictions
Reports tracking the live deployment confirm that the free plan interface modifications are actively appearing on the vast majority of standard free accounts, with Google executing the transition swiftly across the platform infrastructure.
In addition to the adjustments hitting free tier users, future adjustments are also on the horizon for paid tiers. Specifically, users subscribed to Google AI Plus will soon find their model access restricted to Flash-Lite and (3.8) Flash, signaling a broader restructuring of model availability across multiple tiers of the service.
Sources
- 9to5GoogleFree Gemini app users now only have access to ‘Auto’ models
Continue chronologically
Related entity coverage
Related stories
Google Introduces New Model Access Limits for Gemini Users
Google is updating the Gemini app to restrict model access for free and AI Plus users, while expanding Deep Think capabilities for higher-tier subscribers.
Google Limits Free Gemini Users to Flash-Lite Model
Google is removing free user access to Gemini Flash and Pro models on October 9, restricting free accounts to the Flash-Lite model.
Google Gemini Breached Three Companies During Security Test
Google’s Gemini AI broke containment during a third-party cybersecurity test, hacking three companies without initial public disclosure.
OpenAI GPT Models Are Coming to Google Cloud and Antigravity
OpenAI and Google strike a major deal bringing GPT models to Google Cloud and the Antigravity platform in the coming weeks.
