Grok Voice 2.0 at a glance
Grok Voice 2.0 improves reasoning, tool use, and transcription accuracy over Grok Voice 1.0, uses fewer reasoning tokens for faster tool calls, and became the grok-voice-latest default on August 5, 2026.
What changed with Grok Voice 2.0
On July 29, 2026, xAI announced Grok Voice 2.0, the successor to Grok Voice 1.0 from April, for its Speech to Speech API. xAI says it reasons while speaking, uses about 0.4 times the reasoning tokens of its predecessor so tool calls usually run before the end of the first sentence, and was trained to speak in shorter sentences and ask one question at a time.
At launch, xAI cited Artificial Analysis results showing 82.9% on the Speech-to-Speech Quality Index, versus 75.7% for Grok Voice 1.0, and 56.5% on the agentic τ-voice Bench. It also reported a 1.4 times transcription accuracy gain over Grok Voice 1.0 across 24 languages. Pricing is $0.08 per minute of audio, and the grok-voice-latest alias moved to this model on August 5, 2026.
Grok Voice 2.0 in numbers
Values are copied from the linked source on the date shown. They are not IA Compa measurements.
- Speech to Speech API price per minuteas of · Source for Speech to Speech API price per minute: x.ai (opens in a new tab)
- $0.08/min
Why Grok Voice 2.0 matters for businesses
For businesses running voice agents on the alias, the switch happened automatically, which is convenient but can change call behavior. Pin a model version for production phone lines, retest booking and data-capture flows, and budget at the per-minute rate, which is higher than the API's original $0.05.
Put it to work: AI Voice Agents
Sources for this story
- Official
- Grok Voice 1.0 (opens in a new tab)Official
xAI · x.ai
- Release notes (opens in a new tab)Official
xAI Docs · docs.x.ai
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #voice-agents
- #xai
- #grok-voice
- #speech-to-speech
- #pricing


