Grok Voice 3.0
xAI’s speech-to-speech voice model, under 1s latency
Start new thread

Grok Voice 3.0 - xAI’s speech-to-speech voice model, under 1s latency

by•
Grok Voice 3.0 is xAI’s speech-to-speech model for voice agents, which xAI calls the world’s most intelligent. It ranks #1 on the Artificial Analysis Speech-to-Speech Index at 86.9, with latency under one second per xAI. It powers voice in Tesla vehicles and Starlink support calls. Available through the Voice Agent API at $0.08 per minute, same as Grok Voice 2.0. Agents on grok-voice-latest switched to 3.0 automatically.

Add a comment

Replies

Best
Hunter
📌

Grok Voice 3.0 is xAI’s new speech-to-speech model for voice agents. xAI announced it October 14, 2026, and calls it “the world’s most intelligent speech-to-speech model.”

Problem: Voice agents for support calls or in-car assistants have to feel natural. They need to answer fast, handle interruptions and pauses, and still get the task done. A slow or error-prone voice agent loses the caller.

Solution: Grok Voice 3.0 takes speech in and gives speech back directly, without a separate text step. Per xAI, latency is under one second. It handles agentic tasks like verifying customers, troubleshooting hardware, issuing credits and shipping equipment.

How it performs:

  • Speech-to-Speech Index: 86.9, ranked #1, against 83.2 for GPT-Live-1 and 82.6 for Gemini 3.8 Live

  • Big Bench Audio (speech reasoning): 98.5%

  • Task success rate: 94.8%

  • Full Duplex Bench (turn-taking and interruptions): 97.0%

  • Cost per task on Big Bench Audio: $0.025, against $0.030 for GPT-Live-1 and $0.055 for GPT-Realtime-2.1

In use:

  • Tesla vehicles: millions of conversations a day, from checking calendars to ordering coffee, hands-free. It also powers Grok Bot’s voice mode.

  • Starlink call centers: 20,000 customer calls a day in multiple languages, with about 70% resolved without a human

Pricing and access:

  • Voice Agent API: $0.08 per minute ($4.80 per hour), the same price as Grok Voice 2.0

  • Agents using grok-voice-latest switched to 3.0 automatically, with no changes needed

Who it’s for: developers building voice agents, such as support lines and in-car or device assistants, who want a single speech-to-speech model through an API.

· ·