Key points
- Free Gemini users can no longer manually select specific models, relying instead on an "Auto" setting.
- Google introduces three effort levels to replace the previous Extended option for all users.
- The update routes most prompts to Flash-Lite, with complex queries directed to Flash or Pro.
Google is implementing changes to the Gemini application that remove manual model selection for users on the free plan. The company is simultaneously introducing adjustable thinking levels, referred to as effort levels, across its user base.
What happened
Google announced last week that it would alter the Gemini app experience for individuals without a paid subscription. According to 9to5Google, the rollout is currently active on the vast majority of free accounts. Users who previously could choose specific models now see only one option in the picker: "Auto" with "Gemini 3 models."
The company’s support documentation previously identified the default model for this tier as (3.5) Flash-Lite. Google states that the Auto feature directs each prompt to the model best suited for the task. Most requests are processed by Flash-Lite, which the vendor describes as its fastest model.
Prompts requiring deeper reasoning may be routed to the Flash or Pro models. If a user disables smart model selection, the system will use Flash-Lite for all responses. This change means free users cannot manually switch models.
Why it matters
The shift impacts how users interact with Google’s AI tools. The automatic routing means that the underlying engine processing a query can vary based on Google’s assessment of complexity.
This variability reflects a change in user control over model selection. For business leaders, the distinction between free and paid tiers involves different model access. Google AI Plus users will soon be limited to Flash-Lite and (3.8) Flash, indicating a tiered approach to model access.
The introduction of effort levels replaces the previous Extended option. These three new choices align with thinking levels available in Google AI Studio and Antigravity. This standardisation suggests Google is unifying its interface across different products.
What happens next
- Monitor internal usage of Gemini to identify workflows dependent on specific model versions.
- Review access controls to ensure free accounts are not used for critical operational tasks.
- Evaluate whether paid subscriptions are necessary for teams requiring consistent model performance.
- Assess the impact of variable model routing on data handling and compliance requirements.
What to do and how to stay safe: Google
- Audit current Gemini usage within your organisation to determine reliance on specific model capabilities.
- Implement clear policies regarding the use of free-tier AI tools for sensitive or critical business functions.
- Monitor vendor documentation for further changes to model availability and routing logic.
- Ensure that data governance protocols account for the possibility of automatic model selection.
Step-by-step guide: Shadow AI usage: how to control unmanaged model access in your firm
General security guidance from the AppSec Scoop newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Can free Gemini users still choose specific models?
No, free users without a plan can no longer manually switch models and must use the Auto setting.
What model handles most prompts for free users?
Most prompts are sent to Flash-Lite, which Google describes as its fastest model.
What replaces the Extended option in Gemini?
Google is rolling out three new effort levels that align with thinking levels in other Google AI products.
