Google is working on a "major" update for Gemini Live, and what has been revealed so far points to a more advanced experience: a new "Thinking Mode" and an "Experimental Features" section to test features before they are released to everyone.

What Google is preparing for Gemini Live

The clue comes from a teardown (basically, a review of the app's code to detect hidden features) that found clear references to significant changes within Gemini Live. This analysis includes direct mentions of Thinking Mode and an Experimental Features section, suggesting that Google wants to separate the "stable" experience from what's in testing.

This fits with Google's typical strategy: launch new features first as experiments, measure how people respond, and then activate them at scale. In other words, Gemini Live isn't just going to "talk pretty," it also wants to become more flexible, with response modes and tools that users can better control.

Thinking Mode: More Thoughtful Responses

The name says it all: Thinking Mode sounds like a mode where Gemini takes more time (and probably more "calculation") to provide better responses to complex tasks. This type of approach is often useful when requesting something that requires several steps—for example, comparing options, planning something with constraints, or solving a nuanced technical question—instead of responding "on the fly."

At the product level, this also suggests that Gemini Live wants to adapt to two conversation styles: the quick one for everyday use, and the more "in-depth" one for when you're in full productivity mode. And if Google integrates it well, it could be the typical switch that becomes indispensable: "answer me now" vs. "think it over and give me the best version."

Experimental Features: a lab within Live

The other major new feature detected is Experimental Features,which sounds like a container for enabling (or disabling) features under development within Gemini Live.This is relevant because, when an AI is in real-time conversation mode, any change in behavior is immediately noticeable: voice, latency, response style, etc. In practical terms, an "experiments" hub would allow Google to test improvements without compromising all users, and let the most curious users activate new features first.