Gemini could soon stop wasting your time on basic questions, and that's great news

With the release of the Gemini 3 generation of AI models, Google’s smart assistant has racked up brilliant deep-thinking capabilities. However, the lack of a smart automatic model selector actively switching between fast and deep-thinking models — as ChatGPT offers — can result in Gemini taking longer and consuming more tokens for straightforward queries that don’t truly require deep thinking.

An upcoming update to Gemini, however, could fix that, allowing users to manually override thinking and dial back to a quicker response.

Google has been spotted testing a new control that will allow users to stop Gemini from thinking and switch to Flash models for faster responses. We’ve received a tip about a Skip button that appears when Gemini starts thinking while responding. Tapping the button stops it from thinking and switches it to the model with fewer parameters, yielding an instant answer.

Besides eliminating the thinking process, the Skip button ensures that Gemini attempts to answer the question directly, rather than listing out its reasoning. Since it can affect the quality of responses in scenarios that require active thinking, the Skip feature is best suited to factual or closed-ended queries, such as general knowledge or direct math.

The tip received from one of our readers suggests that Google has already begun testing the button with certain users. The test appears to be random at the moment, and we’re unsure of the selection criterion. Moreover, the test appears to be linked to the Google ID rather than the device, and the feature is expected to be available across all platforms, including mobile and web.

However, we’ve also come across a code reference hinting at Gemini’s ability to skip deep thinking. We’ve seen the following strings that pertain to the feature:

With that noted, we’re unsure when Google plans to roll this feature out widely, though we hope it does soon. We’ll keep you updated as soon as we learn more.

Thank you, Telegram user viridivn, for the tip!