xAI on April 23 released grok-voice-think-fast-1.0, a new flagship voice model that the company says topped the industry's τ-voice (Tau Voice) Bench at 67.3% — comfortably above Google's Gemini 3.1 Flash Live at 43.8%, xAI's own prior Grok Voice Fast 1.0 at 38.3%, and OpenAI's GPT Realtime 1.5 at 35.3%. The model is generally available through the xAI API and is already running live on at least one large enterprise deployment.
The announcement, made just before the weekend, frames Grok Voice Think Fast 1.0 less as a consumer demo and more as a production tool for the kind of high-stakes, multi-step phone workflows that have so far resisted automation: airline rebookings, telecom billing escalations, retail order changes.
Benchmarks and capabilities
xAI broke out τ-voice Bench performance by vertical. The new model scored 73.7% on telecom tasks (versus 21.1%–40.4% for competing voice models), 66% on airline workflows, and 62.3% on retail. The company attributes the lead to a full-duplex architecture that lets the model process incoming speech and generate a response simultaneously, plus background reasoning that, per xAI, runs "in real time with no impact on response latency."
The model natively supports more than 25 languages and is designed to operate across 28-plus distinct tools, with structured data capture and read-back for items such as addresses, phone numbers, and account IDs. xAI says it has been tested under realistic telephony conditions including background noise, heavy accents, and frequent interruptions.
Starlink as the showcase deployment
The headline customer is SpaceX-affiliated Starlink. xAI says Grok Voice now handles the entire phone sales and customer support operation at +1 (888) GO STARLINK, reporting a 20% sales conversion rate on inbound calls and a 70% autonomous resolution rate for support inquiries — that is, seven in ten calls reportedly close without a human agent in the loop.
That metric, if borne out independently, is the kind of number that voice-AI vendors like Sierra, PolyAI, and Cresta have been chasing for two years. It also gives xAI a real-world reference account at exactly the moment OpenAI's Realtime API and Google's Gemini Live are pushing into the same enterprise telephony territory.
Implications
Grok Voice Think Fast 1.0 lands in a market where voice has quietly become the next frontier for agentic AI. With Starlink as a flagship and a benchmark lead across every τ-voice vertical, xAI is positioning Grok as a credible enterprise voice stack rather than a chatbot novelty — and giving rivals a clear new performance bar to clear.
