xAI ships Grok Voice 3.0, tops a speech benchmark at the same price
Grok Voice 3.0 answers in under a second and ranks first on a third-party speech-to-speech index, but xAI kept the API price exactly where it was.
xAI released Grok Voice 3.0 on 14 October, available immediately in its Voice Agent API. It is a speech-to-speech model, meaning it responds to audio with audio directly, rather than transcribing to text and back.
Latency is under a second, which xAI says is 40% faster than OpenAI’s GPT-Live-1 Astra. On Artificial Analysis’s Speech-to-Speech Index, a third-party benchmark, Grok Voice 3.0 scores 86.9, ahead of GPT-Live-1 and Google’s Gemini 3.8 Live. xAI also reports a lower cost per completed task, $0.025, against $0.030 for GPT-Live-1 and $0.055 for GPT-Realtime-2.1. The API price itself did not move: it stays at $0.08 a minute, the same as Grok Voice 2.0. xAI says the model already handles Starlink’s call centre traffic, around 20,000 calls a day, with 70% resolved without a human.
Why it matters: if you’re running a voice agent on a competing model, this is a benchmark-backed case for switching that doesn’t come with a price increase. Sub-second latency is the threshold where a voice call stops feeling like it’s waiting on a server.
The caveat: the benchmark numbers come from xAI’s own announcement, quoting a third-party index it didn’t run itself, so treat them as a starting point for your own testing rather than a guarantee.