The anticipation for the annual Google I/O 2026 is almost over, and rumors about Google’s future developments continue to surface at a relentless pace.
Among the most notable news in the last hours is a significant new feature for Google’s in-house AI application, Gemini, which seems ready to offer users unprecedented control over their processing capabilities.
According to the outlet 9to5Google, a small number of users have begun to see a new option named “Thinking Level” within the model selection menu.
Until now, people could only choose among various general versions, such as Fast, Reasoning, or Pro without further customization.
More computing power for Google Gemini, here’s how
With the new update, the interaction gains new nuances. In addition to selecting the desired base version, users will have the option to decide how deeply the software should reason to complete a given task.
This new feature is currently manifested when opting for the Fast version, based on Gemini 3 Flash, or when selecting Gemini 3.1 Pro with the Reasoning mode enabled.
Although the rollout of this feature is still in an early and extremely limited phase, those familiar with developer tools know well that Google AI Studio already offers similar controls, allowing you to set deductive power at low, medium, or high levels.
Bringing this flexibility to the consumer-facing application represents a natural step, at a time when tech giants compete to make their virtual assistants increasingly autonomous and sophisticated.
The importance of balancing speed and accuracy
This new feature has practical utility for everyday use. Not all AI requests require the maximum computing power available.
In many cases, you simply want a quick and immediate response, without waiting several seconds for the system to overly analyze a banal text such as a shopping list, as if drafting a university thesis.
Allowing people to calibrate the right balance between the speed of execution and the accuracy of the analysis will make daily use much more versatile and context-appropriate.
Alongside these improvements, the platform is progressively expanding its network of integrations with third-party applications. Currently, the system already talks with very popular services, including GitHub, OpenStax, Spotify and WhatsApp.
Recent leaked support documentation, however, suggests that further collaborations are in development, with widely used tools such as Canva, Instacart and OpenTable ready to join the list.
Although these new connections are not yet active, the timing of these rumors is perfectly aligned with Google I/O approaching. The event traditionally marks the moment when the company outlines the transition of its AI from a simple text-based interlocutor to a fully fledged digital assistant.
