Grok Voice 3.0

Grok Voice 3.0 pairs near-instant speech-to-speech with top-ranked call accuracy — explore the model and spin up voice agents from $0.08 per minute.

Build Live Voice Agents with Grok Voice 3.0
Speak, listen, and answer in real time on the Voice Agent API
AI Video Prompt Generator

Support

Pro AI Tools

Explore elite tools including the Suno AI Music Generator.

placeholder hero

Why Grok Voice 3.0 Leads in Speech AI

A speech-to-speech model built for real phone calls, Grok Voice 3.0 holds the top accuracy score among frontier voice systems while answering in well under a second.

  • Highest Accuracy on Independent Benchmarks
    It sits at the top of Artificial Analysis's Speech-to-Speech Index with a score of 86.9, a blend of speech reasoning, τ-Voice, and blind head-to-head preference.
  • Replies in Under a Second, Both Ways
    Latency stays below one second and the Full Duplex Bench score reaches 97%, so pauses, turn-taking, and interruptions all feel natural.
  • Tested at Tesla and Starlink Volume
    It already handles millions of exchanges every day across Tesla and SpaceX, including 20,000 Starlink support calls daily — roughly seven in ten closed without a person.

How to Put Grok Voice 3.0 to Work

Four practical steps, from opening API access to running multilingual call automation at scale.

Grok Voice 3.0 Capabilities That Matter in Production

The numbers that decide whether a voice agent ships: accuracy, response speed, robustness under noise, and cost per task.

Record Call Resolution, Independently Verified

Grok Voice 3.0 posts a new high for resolving support calls and holds the #1 spot on Artificial Analysis's Speech-to-Speech Index at 86.9.

Turn-Taking That Feels Human

A 97% score on Full Duplex Bench shows how smoothly the model handles pauses, hands the floor back, and absorbs interruptions mid-sentence.

Holds Up Against Accents and Background Noise

τ-Voice runs agents through support calls with varied accents, noisy rooms, and interruptions; Grok Voice 3.0 finishes at 71.6%.

Smart Answers Without the Wait

On τ-Voice it matches GPT-Live-1 (Astra) while running 40% faster, widening how clever and how quick an agent can be at once.

Cheaper per Completed Task

In Artificial Analysis's Big Bench Audio run, spend per task lands under both GPT-Live-1 (Astra, Medium) and GPT-Realtime-2.1.

Built for Enterprise-Grade Accuracy

On ServiceNow's EVA-Bench it takes the top accuracy mark at 74% and the best combined accuracy-plus-experience result.

FAQ

Grok Voice 3.0: Your Questions Answered

Quick answers on accuracy scores, response speed, per-minute pricing, and how the model behaves on live customer calls.

1

What exactly is Grok Voice 3.0?

It is a speech-to-speech model presented as the most intelligent of its kind, and it is available in the API at the same rate as the previous version.

2

How accurate is it, really?

It holds the top position on Artificial Analysis's Speech-to-Speech Index at 86.9 and also scores 97% on Full Duplex Bench.

3

What latency should I expect on live calls?

Under one second, at the front of its class — and roughly 40% quicker than GPT-Live-1 (Astra) on τ-Voice.

4

What does Grok Voice 3.0 charge?

Nothing changed: the Voice Agent API bills $0.08 per minute, or $4.80 per hour, exactly as Grok Voice 2.0 did.

5

Does it work on real support calls?

It does — Starlink contact centers send 20,000 customer calls a day through Grok Voice in several languages, and roughly seven in ten resolve without a human.

6

Will my grok-voice-latest agent need rework?

Not at all. If your agent already targets grok-voice-latest, the upgrade was applied quietly in the background.

Launch Your First Grok Voice 3.0 Agent

Get the top-ranked speech-to-speech model answering calls in under a second, at enterprise scale, with Voice Agent API pricing still at $0.08 per minute.