Grok Voice 3.0 takes first place at the old $0.08 price

On Artificial Analysis' independent Speech-to-Speech Index, the new xAI model scored 86.9 and ranked first, according to xAI as reported by Gate. GPT-Live-1 follows at 83.2, Gemini 3.8 Live at 82.6, and the previous Grok Voice 2.0 at 81.3.

Speech-to-speech means the model hears audio and answers in audio, with no separate text step in the middle. xAI puts end-to-end latency below 1 second, and the model is live in the Voice Agent API at $0.08 per minute, the same as the previous version.

Agents set to grok-voice-latest move to the new model automatically, so nobody has to touch their config.

The SPACEXSI account on Threads adds a claim of 40% lower latency, but Gate's report of the xAI announcement carries neither that figure nor a baseline for it.