Google Gemini: an "Extended" mode for more accurate responses

Google is experimenting with a new feature in its Gemini app, called "Thinking Level." This option allows users to choose between two modes: "Standard" and "Extended." The goal is to adjust the reasoning effort of the artificial intelligence before it provides a response.

The "Extended" mode is designed to increase the number of internal reasoning steps that the AI performs. While this may slow down response time, this approach aims to improve accuracy, particularly for complex tasks that are often prone to AI hallucinations.

Features of the "Thinking Level"

This option is available for models such as Gemini 3 Flash and Gemini 3.1 Pro. By selecting the "Extended" mode, users can expect a reduction in inaccurate responses thanks to a more in-depth reasoning process.

Deployment and Future Integrations

The rollout of this new feature is gradual and is not yet accessible to all users. To date, some users, even those with a Pro subscription, still do not have access to this option in their Gemini apps. Google has not yet officially confirmed the feature or specified the pace or area of deployment.

Meanwhile, the Gemini app plans to integrate more third-party applications, such as Canva, Instacart, and OpenTable. However, these integrations are not yet available.

Context and Expectations

These developments come as the Google I/O 2026 conference approaches, scheduled for May 19. Google is expected to focus on artificial intelligence, with announcements regarding new models, services, and integrations.