SpaceXSI, formerly xAI, released Grok Voice 3.0 on October 14, 2026, bringing its latest speech-to-speech model to the Voice Agent API. The release tops Artificial Analysis' Speech-to-Speech Index while retaining the previous model's $0.08-per-minute price.
On October 14, 2026, SpaceXSI announced Grok Voice 3.0, the new speech-to-speech model behind its Voice Agent API. The company is targeting customer support and sales agents that need to understand callers, respond naturally, and complete tasks during a conversation. Its launch announcement reports response latency under one second and says the model is already powering conversations across Tesla and SpaceX.
Artificial Analysis ranked Grok Voice 3.0 first on its Speech-to-Speech Index with a score of 86.9, ahead of GPT-Live-1 at 83.2 and Gemini 3.8 Live at 82.6, according to the announcement and SiliconANGLE's coverage. The model also scored 97% on Full Duplex Bench, which evaluates pauses, turn-taking, and interruptions. The results vary by task: GPT-Live-1 retained a lead on the support-call benchmark Tau-Voice, scoring 74.5% against Grok's 71.6%.
The release also highlights production deployments. SpaceXSI says Starlink uses Grok Voice to answer about 20,000 customer calls a day across multiple languages, resolving roughly 70% without human intervention. The company describes an agent that can verify customers, troubleshoot hardware, issue credits, and arrange replacement equipment. In Tesla vehicles, it says the model handles millions of daily conversations, including hands-free requests involving calendars, email, and bookings.
Grok Voice 3.0 is available through the Voice Agent API at $0.08 per minute, or $4.80 per hour, matching Grok Voice 2.0 pricing. Agents using the grok-voice-latest alias have already been upgraded automatically. For developers evaluating conversational voice systems, the release provides a new option to assess for responsiveness, interruption handling, and task completion alongside speech quality.
