Gemini’s desktop Tasks mode comes with a voice button

The new Tasks mode in the Gemini desktop app has the same Gemini Live button as regular chat. That points to running Computer Use jobs by voice: you would keep a live conversation going while Gemini operates other apps on your laptop. TestingCatalog found the Live entry point in an app update it reported on September 24.

The update adds a switch between Chat and Tasks, and trusted testers are starting to get access to Tasks. TestingCatalog doesn’t have access yet, so nobody outside the test group has seen how the mode behaves once it’s open. In August the same feature turned up as a “Task” option in the prompt-bar selector, listed next to image and video generation. Now it has a switch of its own, next to Chat.

The voice side relies on Gemini 3.8 Live, the real-time audio model Google announced on September 15 together with 3.8 Live Extended Thinking. Google says it can carry out tasks during a dialogue and supports asynchronous function calling. With asynchronous calls, the model starts a tool call and keeps talking while the call runs, instead of going quiet until the result comes back. TestingCatalog calls the model a natural fit for driving a computer by voice.

The Computer Use settings are also more detailed now. A new menu lists allowed and disallowed apps, so you choose which programs Gemini can operate. The option to start a Gemini project straight from a folder, spotted earlier, is now more prominent and appears to be entering trusted testing as well.

The Windows version of the app launched on September 10 for Windows 10 and 11. It opens over whatever you’re working on with Alt + Space, and it lets you hand multi-step tasks to Gemini Spark, which Google describes as a 24/7 personal AI agent. So the Windows app has been public for only two weeks, and Google is already testing a mode that operates the desktop around it. Spark already takes on multi-step work, so Tasks looks like the step from delegating a job to Gemini working directly in your apps.

There is also a Gemini 4 rumor. Some testers have reported responses that seem to come from a Gemini 4 Pro checkpoint while they were using Gemini 3.8 Flash, TestingCatalog says, but it isn’t clear which model actually answered.

A production rollout of the Computer Use features could still be weeks away or more, depending on how testing goes.