xAI launches Grok Voice 3.0 speech-to-speech model for voice agents

On Oct 14, xAI released Grok Voice 3.0 for the Voice Agent API at $0.08 per minute of input audio, the same price as Grok Voice 2.0. Agents on the grok-voice-latest alias switched over automatically. Artificial Analysis ranks it #1 on its Speech to Speech Index at 86.9, in results it published Oct 14. xAI reports sub-second replies, a 71.6% τ-Voice score it calls comparable to GPT-Live-1 (Astra) at lower latency, use in Tesla vehicles, and Starlink call centers answering about 20,000 calls a day with roughly 70% resolved without a human — those call-center figures are from xAI.
How you’d use it: In the Voice Agent API, point your agent at grok-voice-latest (or Grok Voice 3.0) and compare latency and support-call handling against your current speech-to-speech model.
Sources
- Original source: xAI ↗
- Voice Agent API docs ↗
- Artificial Analysis Speech-to-Speech ↗
- xAI announcement on X ↗