Grok Voice is SpaceXSI's speech-to-speech model: it listens to what you say and replies out loud in real time, so apps can hold spoken conversations. Version 3.0 arrived on 14 October 2026.
In plain words
A speech-to-speech model takes speech in and gives speech back. Developers reach Grok Voice through SpaceXSI's realtime API over WebSockets, a connection that stays open while you talk, and the model can use tools during the call. SpaceXSI says version 3.0 replies in under a second and copes well with pauses and interruptions. In the API it costs $0.08 a minute, the same price as the version before it.
A quick example
Imagine phoning a shop's help line and talking to an AI agent. You ask where your parcel is, it checks with a tool, answers in a natural voice, and lets you cut in halfway through a sentence.
Where you'll meet it
- Voice agents that answer customer support calls.
- Spoken assistants in cars; SpaceXSI says Tesla drivers use it for millions of chats every day.
Related
Sources
- DocsIntroducing Grok Voice 3.0, SpaceXSI · read 14 Oct 2026
- DocsVoice Overview, SpaceXSI Docs · read 14 Oct 2026
Checked by WhatIsAI against the sources above · 14 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.