x.ai released Grok Voice 3.0. It is a speech-to-speech model. It ranks first on the Speech-to-Speech Index with 86.9. It scores 97% on Full Duplex Bench. It has under one second latency. It costs less than comparable models. It is used by Tesla and SpaceX for millions of daily conversations. It resolves 70% of Starlink calls without human agents. It is available now in the API at $0.08 per minute. Existing agents on grok-voice-latest were switched automatically.
What changed
Grok Voice 2.0 was the prior version at the same price.
What it unlocks
Running a voice agent with under one second latency and lower cost per task.
- 86.9 Speech-to-Speech Index score
- 97% Full Duplex Bench score
- $0.08 per minute
- 20,000 daily calls at Starlink
What you need to act on it
- API access
- grok-voice-latest model tag
Sources