A screenshot circulating on X shows a “Start a voice call” control inside Grok Bot, in a build that is not yet live for users. The new button will allow users to start a voice call with their Grok Bot, enabling verbal interaction within the Grok Bot app.
Grok Bot is xAI’s system of persistent AI teammates that run on their own cloud computers. Users currently message those Bots from the desktop and iOS apps, and the Bots then work inside a shared cloud machine. A voice call would sit on top of that teammate model, not replace it.
Voice Commands
The image shows a desktop chat with a Bot named Test Bot demonstrating the upcoming templates feature, but more importantly, has a headset icon with a tooltipin the top right corner, which says “Start a voice call.”
SpaceXAI’s team seems to be working on providing the ability to start voice calls with your Grok Bots, giving users a quick and simple way to interact, similar to other voice-controlled AI services. For now, there doesn’t appear to be an associated wake word for mobile or active microphone monitoring, which suggests that Grok Bot’s Voice Call will be a user-initiated action at the start.
In addition, the composer’s microphone button is still visible in the bottom right, indicating that for more precision and for using the main thinking models that Grok Bot uses, users can still opt to dictate into the chat instead of holding a back-and-forth conversation.
Likely Using Grok Voice Think Fast 2.0
xAI already has a realtime speech-to-speech stack. On 29 July 2026, the company announced Grok Voice Think Fast 2.0 as its most capable speech-to-speech voice model. xAI’s comparison, citing Artificial Analysis, lists a 0.70-second time to first audio, down from 1.25 seconds on Think Fast 1.0.
Official developer docs now describe grok-voice-think-fast-2.0 as the flagship voice model, with grok-voice-latest as an alias for that release. The company’s Voice pages also mention a Voice Agent Builder for phone and in-browser agents. Think Fast 2.0 is priced at $0.08 per minute of audio.
So far, SpaceXAI has not confirmed which model will sit behind a Grok Bot voice call, or whether the feature will use an entirely new stack.
What Still Has to Happen
A tooltip and button in an upcoming build is a sign that implementation is around the corner, but not quite here yet. Grok Bot still has to expose the button in a live customer build, and complete a two-way audio session in which a Bot does useful work.
Until SpaceXAI confirms the feature, we’ll still be waiting to see when and on which platforms this launches.
