Google is rolling out an update to the Gemini app for free-tier users, shifting them to automated model selection while introducing new thinking effort levels to replace the previous option, 9to5google.com reported.
Free Tier Restricted to Flash-Lite Under Auto Model Selection
Users without a paid AI subscription can no longer manually switch between different Gemini models in the app picker. Their interface now defaults strictly to an “Auto” setting powered by Gemini 3 models, which support documentation identified as Flash-Lite. Under this routing system, the app evaluates incoming prompts automatically.
Most routine prompts route directly to Flash-Lite, which Google designates as its fastest architecture. Prompts requiring deeper reasoning scale upward to the Flash or Pro models. When users turn off smart model selection manually, chat interfaces for non-subscribers route all responses exclusively through Flash-Lite.
According to reporting by SammyGuru, regular Flash and Pro models are disappearing entirely from the free tier. Subscribers using Google AI Plus retain access to Flash-Lite and the 3.8 Flash model, while higher tiers like AI Pro and AI Ultra unlock the full suite including the Pro architecture.
Granular Effort Levels Replace Extended Thinking Options
Alongside model restrictions, Google is rolling out three new thinking or “effort levels” across the application ecosystem. These configurations supersede the older “Extended” setting. Users can select between Low for quick tasks, Medium for balanced analytical depth, and High for extra thorough processing.
These effort parameters mirror the thinking levels already present in Google AI Studio and Antigravity environments. Subscribers to Google AI Pro also gain access to Deep Think, a feature that previously remained exclusive to AI Ultra accounts. Deep Think utilizes parallel processing architectures to evaluate multiple analytical paths before generating a final response.
| Subscription Tier | Available Models | Usage Limit Multiplier | Deep Think Access |
|---|---|---|---|
| Free Plan | Flash-Lite (Auto-routed) | Standard (1x) | No |
| Google AI Plus | Flash-Lite, (3.8) Flash | 2x Standard | No |
| Google AI Pro | Flash-Lite, Flash, Pro | 4x Standard | Yes |
| Google AI Ultra | Flash-Lite, Flash, Pro | 5x to 20x Pro | Yes |
Usage Quotas and Platform Rollout Pace
The transition for free accounts is deploying rapidly across user bases, with widespread availability observed on accounts evening Pacific Time. While model access tightens for non-paying users, structural usage limits remain tied to subscription tiers rather than these new routing mechanics.
Free accounts retain standard usage allowances. Google AI Plus doubles that baseline, AI Pro quadruples it, and AI Ultra scales from five to twenty times the AI Pro limit depending on the exact tier configuration. Selecting higher effort levels consumes a larger portion of these localized usage quotas during complex task execution.
