xAI Launches Grok Voice 3.0, Keeping API Pricing at $0.08 a Minute
xAI says its new speech-to-speech model leads independent benchmarks, answers in under a second, and already runs in Tesla cars and Starlink support lines.
xAI has announced Grok Voice 3.0, which it calls “the world’s most intelligent speech-to-speech model.” It’s available now in xAI’s Speech to Speech API.
Price
The headline for developers is that the price hasn’t moved. Grok Voice 3.0 costs $0.08 per minute ($4.80 per hour), the same as Grok Voice 2.0. xAI says apps that call grok-voice-latest have already been moved to the new model with no code changes.
The benchmark claims
xAI cites Artificial Analysis testing dated October 13, 2026:
- 86.9 on the Artificial Analysis Speech-to-Speech Index, up from 81.3 for Grok Voice 2.0. xAI lists GPT-Live-1 at 83.2 and Gemini 3.8 Live at 82.6.
- 97% on Full Duplex Bench, which measures pauses, turn-taking and interruptions.
- 71.6% on τ-Voice, a support-call test with accents, noise and interruptions. xAI says that’s comparable with GPT-Live-1 (Astra) at 40% lower latency.
- 74% accuracy on ServiceNow’s EVA-Bench for enterprise voice agents, the top score according to xAI.
xAI also says response latency is under one second. These are xAI’s figures, citing Artificial Analysis and ServiceNow. We haven’t run our own tests.
Where it’s already running
According to xAI, Grok Voice handles millions of conversations a day in Tesla vehicles and powers Grok Bot’s voice mode. xAI says Starlink call centers use it for about 20,000 calls a day, with roughly 70% resolved without a human.
Source: xAI, “Introducing Grok Voice 3.0”.
