Google is once again caught in a wave of pre-I/O leaks, and the latest find inside the Gemini app suggests the company is experimenting with giving users finer control over how much cognitive effort the AI invests before delivering an answer. A new &8220;Thinking Level&8221; option has reportedly appeared in the app&8217;s model picker, where users currently choose between Fast, Thinking, Pro, or Google AI Plus. The feature, which mirrors controls already present in Google AI Studio, would allow users to adjust the reasoning depth of certain models on the fly.

What is the Thinking Level feature?

According to reports from several users, the Thinking Level option currently appears when selecting the Fast (Gemini 3 Flash) model or Gemini 3.1 Pro with thinking enabled. The integration is still clearly experimental, with the rollout extremely limited at this stage. However, its existence points to a broader trend in the AI landscape: companies are now competing not just on raw intelligence but on the granularity of user control over that intelligence.

In Google AI Studio, developers already have access to Low, Medium, and High reasoning settings for certain models. Bringing this same flexibility to the consumer-facing Gemini app would be a logical next step. It would allow end users to decide whether they want a lightning-fast answer to a simple fact-check or a deeply reasoned response to a complex analytical question. The concept itself is not new &8211; OpenAI&8217;s ChatGPT has offered something similar through its &8220;temperature&8221; and &8220;top-p&8221; parameters in the API, but those remain largely invisible to the average user. Google seems intent on making such controls more accessible and intuitive.

Why adjustable reasoning matters

Not every AI request demands maximum cognitive horsepower. Asking for the weather, a quick definition, or a reminder does not require the model to simulate multiple hypothetical scenarios or weigh moral implications. Yet many of today&8217;s large language models default to a one-size-fits-all reasoning budget, often resulting in overly verbose or slow responses for trivial queries. The Thinking Level feature aims to solve that by letting users dial down the reasoning for everyday tasks and dial it up for research, brainstorming, or decision-making.

From a technical standpoint, adjusting the reasoning level likely controls the number of internal reasoning steps the model takes before generating a response. In many transformer-based models, deeper reasoning involves multiple passes or chain-of-thought processing, which increases latency and computational cost. By exposing this control, Google can help users balance speed against depth, potentially also affecting the cost of API calls for developers when similar controls make it into the platform.

The move also aligns with the industry&8217;s push toward agentic AI. As assistants become more capable of executing multi-step tasks, they need to know when to think deeply and when to act quickly. A Thinking Level slider could be the first step toward more adaptive AI behavior, where the model itself learns to adjust its reasoning based on the complexity of the query, with user override capability.

What this means for the future of Gemini

Google&8217;s broader strategy appears to be transforming Gemini from a simple chatbot into a comprehensive digital assistant that operates across apps and services. The Thinking Level feature fits neatly into that vision: it empowers users to tailor the AI&8217;s behavior to their current context, whether they are quickly checking facts or engaging in a deep creative session.

At the same time, Google is expanding Gemini&8217;s integration ecosystem. The app already works with major platforms like GitHub, Spotify, and WhatsApp, and documentation suggests integrations with Canva, Instacart, and OpenTable are on the horizon. These integrations, combined with adjustable reasoning, would allow Gemini to handle everything from grocery list preparation to graphic design tasks with appropriate levels of cognitive investment.

The timing of these developments is telling. As other AI companies race to release ever-larger models with more parameters, Google is focusing on practicality and user experience. The Thinking Level feature acknowledges that bigger is not always better &8211; sometimes a smarter AI is one that knows when not to overthink. By giving users control, Google hopes to make Gemini feel less like a monolithic oracle and more like a helpful assistant that adapts to your needs.

Industry analysts have pointed out that such customization could also reduce costs for users who are billed per query or per token. Lower reasoning levels mean fewer tokens processed, which translates directly to savings for power users. For free-tier users, it could mean faster responses and less frustration with AI that refuses to give a simple answer without a lengthy preamble.

Critics might argue that exposing reasoning controls adds unnecessary complexity to an interface that should remain simple. However, early user feedback from Google AI Studio suggests that developers appreciate the flexibility, and similar controls have been well-received in other AI tools. The challenge for Google will be to design the consumer-facing implementation in a way that feels natural and does not overwhelm new users.

Looking ahead, the Thinking Level feature is likely just the beginning. Google may eventually introduce automatic reasoning adjustment, where the AI determines the optimal level based on the query&8217;s complexity and the user&8217;s historical preferences. This would create a seamless experience where the assistant is always just smart enough and never too slow.

As with all unreleased features, there is no guarantee that Thinking Level will see a public rollout. Google frequently tests features that never ship, or that take months to reach a wide audience. Still, the fact that it appears in a production version of the app, even in a limited form, suggests the company is serious about making it available.

In the meantime, Gemini users can look forward to a more responsive and customizable AI experience. The days of one-size-fits-all chatbot responses are fading, replaced by assistants that respect both your time and your need for depth. With the Thinking Level feature, Google is betting that the smartest AI is the one that knows exactly how much thinking you need it to do.

Source: Digital Trends News