xAI Grok Voice 3.0 Ships at $0.08/Min, Tops Index at 86.9
TL;DR
- Grok Voice 3.0 tops the Artificial Analysis Speech-to-Speech Index at 86.9 while pricing at $0.08/min, 17-55% below comparable OpenAI offerings.
- Starlink routes 20,000 daily customer calls through Grok Voice 3.0, with 70% resolved end-to-end without a human agent.
- One in four Tesla sales calls handled by Grok Voice 3.0 results in a purchase before the caller disconnects.
xAI released Grok Voice 3.0 on October 14 at $0.08 per minute, holding the price flat against the prior version while claiming the top slot on Artificial Analysis's Speech-to-Speech Index at 86.9. The company's launch post puts the model at 97% on Full Duplex Bench, 71.6% on τ-Voice and 74% on EVA-Bench, with sub-second response latency it describes as 40% lower than GPT-Live-1 (Astra).
The deployment story is where xAI is pushing hardest. The company says the model already handles millions of daily conversations through Tesla's in-car assistant and Starlink's support line, with Starlink alone taking roughly 20,000 customer calls a day and resolving about 70% of them without a human. Sales calls, by xAI's own figure, convert at 25%.
The one outside voice carried in the launch is from ServiceNow. "Grok Voice 3.0 sets a new Pareto frontier on EVA-Bench," said Nitin Aggarwal, ServiceNow's Senior Director of Product Management for Agentic Evals and RAI, adding that it "pairs top accuracy with a natural, responsive conversation."
The pricing is doing most of the work in the pitch: $0.025 per task on Big Bench Audio versus GPT-Live-1 at $0.030, with the per-minute rate unchanged from Grok Voice 2.0. Every benchmark figure in the post traces back to xAI or Artificial Analysis; no independent audit of the Starlink or Tesla deployment numbers accompanies the release.
This is the 49th xAI item in our 90-day tracker, landing in a voice-AI cycle that has already produced 41 others in the same window.
What others are reporting
-
KuCoin / MetaEra Read →
Only source to flag the governance gap: no model card, no security audit, and a live system executing refunds and shipments. Also documents a latency regression versus Grok Voice 2.0 and a customer-service benchmark deficit behind GPT-Live-1.
The voice model is becoming an Agent entry point that can directly interact with business systems.
-
Tesla North Read →
Grounds the benchmarks in live Starlink call-center data, reporting 20,000 daily calls at 70% autonomous resolution, plus Tesla in-car deployment at millions of conversations per day.
roughly 70% of them are resolved without a human stepping in, while one in four sales calls ends in a purchase
-
Gagadget Read →
Adds competitive pricing math, calculating a 17-55% undercut versus OpenAI, and frames Tesla in-car voice as a near-term interface for autonomous driving.
The model already processes millions of conversations daily in Tesla cars.
-
Basenor Read →
Frames the launch as vertical integration proof: one model spanning enterprise B2C support at Starlink and consumer in-car experience at Tesla, with notes on potential FSD voice control.
20,000 customer calls per day, with roughly 70% resolved end-to-end without a human agent ever picking up.
Originally reported by x.ai
Read the original article →Original headline: xAI Ships Grok Voice 3.0 at $0.08/Min, Tops Speech-to-Speech Index at 86.9 Running Starlink and Tesla Calls