ModelCards

Grok Voice 3.0

Audio modelCard summarized

The world’s most intelligent voice model.

Source: Introducing Grok Voice 3.0 (launch post) and Speech to Speech docs; no model card published

Use it via API

$0.08 / min
Code exampleCompare

Where to call it

  • xAI APIgrok-voice-3.0

Supports

  • Web search
Full feature support

Good for

  • Customer-support and sales phone agents (SIP, native G.711)
  • In-car and device voice assistants
  • Voice agents that call tools mid-conversation (search, MCP, functions)

Watch out for

Overview

xAI · 02Voice, front

Grok Voice 3.0

The world’s most intelligent voice model.

Typical price
$0.08 / min
Speech to Speech session
Max length
—
Languages
20 listed
Modes
speech-to-speech, voice agents with tool calling
Released
14 Oct 2026

Grok Voice 3.0 is xAI's speech-to-speech model for real-time voice agents, served through the Speech to Speech (realtime WebSocket) API at the same price and with the same API as Grok Voice 2.0. xAI says it ranks first on Artificial Analysis's Speech-to-Speech Index (86.9), responds in under one second, and handles millions of conversations a day in Tesla vehicles and Starlink call centers. launch post; Speech to Speech docs; release notes

grok-voice-3.0
Typical price
$0.08 / min
Languages
20 listed (incl. Arabic, Chinese, Hindi, Japanese, Spanish and Portuguese variants); more with varying accuracy
Weights
Closed
Released
14 Oct 2026
  • speech-to-speech
  • voice agents with tool calling
  • custom voices (cloning)
  • SIP phone calls
  • OpenAI Realtime API compatible