DEDUPEDElon Musk's AI company, formerly xAI, released Grok Voice 3.0, the speech-to-speech model behind its Voice Agent API. It listens, reasons, calls tools and talks back, and it is pitched squarely at businesses running voice agents for customer support and sales calls.
On the Artificial Analysis Speech-to-Speech Index it ranks first with 86.9, ahead of GPT-Live-1 at 83.2 and Google's Gemini 3.8 Live. The company says responses come back in under a second and that it set a new record for resolving support calls. The price is unchanged from version 2.0 at $0.08 a minute, about $4.80 an hour of input audio, slightly above GPT-Live-1 and Gemini Live.
It does not win everywhere. On the separate tau-Voice benchmark it scores 71.6%, behind GPT-Live-1 at 74.5%. It is also hosted only: there are no open weights or local versions. The company says Grok Voice already handles millions of conversations a day inside Tesla vehicles, which gives it plenty of real-world traffic to learn from.
Our take
Voice agents have quietly become a three-way race, and leaderboard leads are now measured in a few points. Price and latency will decide more deployments than the index.
From the issue
Anthropic is profitable and heading to IPO, and the sue-everyone era just began Thursday, October 15, 2026