Grok Voice 3.0 Launches: Faster Voice AI at the Same $0.08 a Minute

Independent review: we don't accept payment for reviews, scores or rankings. How we test

Published October 15, 2026 · AI News

SpaceXSI, the AI company formerly called xAI, released Grok Voice 3.0 on October 14, 2026. It is the speech-to-speech model behind Grok’s voice agents, and it keeps the same API price as the last version: US$0.08 a minute.

The new model now powers Grok Bot’s voice mode and runs in Tesla cars. Developers using the grok-voice-latest alias were moved over automatically.

What changed

Grok Voice 3.0
API priceUS$0.08/min (US$4.80/hour), unchanged
Approx. in CADabout CA$0.11/min (CA$6.53/hour) at 1.36
Response timeUnder one second, per SpaceXSI
Artificial Analysis Speech-to-Speech Index86.9 (#1), vs 83.2 for OpenAI GPT-Live-1 and 82.6 for Gemini 3.8 Live
Tau-Voice (task success)71.6%, just behind OpenAI’s 74.5%

SpaceXSI also points to real use: its Starlink voice agent handles about 20,000 customer calls a day in several languages, and the company says roughly 70% are resolved without a human. Those are company figures, not independent audits.

What it means for you

If you just talk to Grok on your phone, the change is mostly under the hood: faster replies and better handling when you interrupt. The bigger story is for businesses building phone agents.

  • For small businesses in Canada and the U.S.: talk time costs about CA$6.50 an hour before tax, so a pilot is cheap to run. Budget for GST/HST on top if you are billed as a Canadian business.
  • For developers: benchmark wins are vendor-reported. Test it on your own calls, accents and background noise before switching from OpenAI or Google.
  • French support: SpaceXSI says it handles multiple languages but does not publish a list. Quebec businesses should test French-language calls first.

Our take: voice AI pricing is now a race to the bottom, and that is good for buyers. Quality on messy, real-world calls is still what decides it.

Looking for a voice tool for videos or narration instead of phone agents? See our best AI voice generator guide.

Sources