Google is giving Gemini users more control over how much reasoning the AI assistant puts into its answers, introducing three thinking levels designed for different kinds of tasks. The update replaces the previous Extended Thinking option with Low, Medium, and High settings, making it easier to balance response speed against more thorough reasoning.

As reported by 9to5Google, the new thinking levels are rolling out alongside a significant change to the Gemini app’s free experience: users without a paid Google AI subscription can no longer manually select individual models and instead see an Auto option.

The new controls are intended to give users a straightforward way to choose how much effort Gemini spends on a response, rather than relying on a single extended-reasoning setting.

What the three Gemini thinking levels do

The new settings offer three levels of reasoning effort, each suited to a different type of request.

• Low: Prioritizes quick, efficient responses, making it suitable for straightforward questions and everyday tasks.
• Medium: Balances response speed with reasoning depth for more involved questions.
• High: Allows for more thorough reasoning, making it better suited to complex problems that benefit from additional analysis.

The options bring Gemini’s consumer app closer to the controls already available in Google’s AI Studio and Antigravity development environment.

The practical difference is that users can adjust the level of effort according to the task. A simple factual question generally does not need the same degree of reasoning as a complicated coding problem or a multi-step analysis.

However, selecting a higher thinking level does not guarantee a correct answer. It gives the model more room to work through a problem, but users should still verify important information.

Free Gemini users now get an Auto model picker

The thinking-level update arrives as Google changes how people on the free Gemini plan access its AI models.

Previously, free users could manually switch between available models. Under the new arrangement, the model picker displays Auto, which lets Gemini choose the model for each prompt. Google says most requests will go to Flash-Lite, while prompts that need deeper reasoning may be routed to Flash or Pro. If automatic model selection is disabled, free users will receive responses from Flash-Lite instead.

That distinction matters because the Auto option does not mean every prompt gets the same model. Google is attempting to match the model to the request, although the company says most free-tier prompts will use its faster Flash-Lite model.

The rollout also comes as Google tightens model access for its entry-level subscription. According to 9to5Google, AI Plus users will eventually be limited to Flash-Lite and Flash, while access to Pro remains available through higher subscription tiers.

Higher thinking levels can use more of your allowance

More reasoning can come at a cost. Google’s Gemini documentation explains that advanced models and higher thinking levels consume more usage, meaning users may reach their limits sooner when tackling demanding tasks.

For everyday users, Low should be sufficient for quick questions, while Medium offers a reasonable middle ground. High makes more sense when a task calls for careful analysis, although it may take longer and consume more of the available usage allowance.

The three new settings are rolling out across Gemini, with availability potentially varying between accounts as Google expands access. The update gives users a more granular way to manage reasoning effort, even as Google’s model-selection changes make manual model choice less accessible to people using the free plan.