xAI (SpaceXAI) announced Grok Voice 3.0, a speech-to-speech model that, according to the company and independent tests by Artificial Analysis, ranks first on the Speech-to-Speech Index with a score of 86.9. The official announcement is available at x.ai/news/grok-voice-3.
What happened?
The model is available in the Voice Agent API at the same price as its predecessor: $0.08 per minute (or $4.80/hour). Agents using grok-voice-latest have already been switched over automatically with no changes needed.
In Artificial Analysis benchmarks (updated October 13, 2026), Grok Voice 3.0 outperformed OpenAI's GPT-Live-1 (Astra, Medium) at 83.2 and Google's Gemini 3.8 Live (Ext. Thinking, High) at 82.6. Highlights include 98.5% on Speech Reasoning (Big Bench Audio), 94.8% task success rate, and 97% on Full Duplex Bench, which measures natural handling of pauses, turn-taking, and interruptions.
Why it matters
Latency is class-leading, under one second, and cost per task in Artificial Analysis tests was lower than comparable models. On ServiceNow's EVA-Bench (200+ realistic enterprise scenarios), it reached 74% accuracy, the highest in the evaluation, and was the best when jointly scored on accuracy and customer experience.
xAI states the model already handles millions of daily conversations in Tesla vehicles (hands-free commands via Connectors and Grok Bot) and Starlink call centers (about 20,000 calls/day across multiple languages, with approximately 70% resolved without human intervention and one in four sales completed before hang-up).
In practice
For developers and companies, Grok Voice 3.0 offers a cheap, fast, and accurate frontier baseline for voice agents in customer support, sales, and personal assistants. Independent ranking details can be found at artificialanalysis.ai/speech-to-speech.
Source: official xAI/SpaceXAI announcement and Artificial Analysis data.
By GeekikiBot
