Grok Voice 3.0
Grok Voice 3.0 delivers sub-second, full-duplex speech-to-speech for voice agents, with top-ranked accuracy and API access from $0.08 per minute.
Support
Pro AI Tools
Explore elite tools including the Suno AI Music Generator.
MiniMax H3
MiniMax H3 AI Video Generator
Seedance 2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
Veo3.1
Create Stunning Videos with Veo3.1
FLUX 3 Video Generator
Kling 3.0
Next-Gen AI Video Generator
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
Grok Voice 3.0: Engineered for Real Conversations
Ranked first for speech-to-speech accuracy, this full-duplex model answers live calls in under a second and keeps enterprise support lines running at scale.
- Highest Accuracy on Independent BenchmarksScores 86.9 on Artificial Analysis's Speech-to-Speech Index, a ranking that blends speech reasoning, τ-Voice, and blind head-to-head preference votes.
- Conversations That Feel InstantLatency stays below one second, and a 97% Full Duplex Bench score means pauses, turn-taking, and interruptions all behave the way people expect.
- Tested on Millions of Real CallsIt already handles millions of conversations daily across Tesla and SpaceX, including 20,000 Starlink support calls a day, with roughly 70% resolved without a human.
How to Put Grok Voice 3.0 to Work
Four practical steps, from getting API access to running multilingual call automation at scale.
Capabilities That Matter in Production
The benchmarks worth tracking for live voice agents: accuracy, latency, robustness, and cost per completed task.
Record Call Resolution, Independently Verified
Grok Voice 3.0 resolves more support calls than any model before it and holds first place on Artificial Analysis's Speech-to-Speech Index at 86.9.
Turn-Taking That Sounds Human
A 97% Full Duplex Bench score shows how cleanly the model handles pauses, speaker switches, and interruptions during live conversations.
Holds Up Under Accents and Background Noise
τ-Voice stress-tests agents with varied accents, noisy rooms, and interruptions; this model reaches 71.6%.
Smarter Answers Without the Wait
It matches GPT-Live-1 (Astra) on τ-Voice while running about 40% faster, widening what an agent can be smart enough — and quick enough — to do.
Spend Less per Completed Task
In the Big Bench Audio evaluation from Artificial Analysis, per-task cost lands below both GPT-Live-1 (Astra, Medium) and GPT-Realtime-2.1.
Strongest Enterprise Call Accuracy
On ServiceNow's EVA-Bench it posts the top accuracy score of 74% and the best combined accuracy plus customer-experience result.
Grok Voice 3.0: Your Questions Answered
Quick answers on accuracy, latency, pricing, and how the model performs on real customer calls.
What does Grok Voice 3.0 actually do?
It is a speech-to-speech model positioned as the most intelligent of its kind, and it is live in the API at the same rate as the model before it.
How well does it score on accuracy?
First place on Artificial Analysis's Speech-to-Speech Index at 86.9, plus a 97% result on Full Duplex Bench.
Is it fast enough for real-time calls?
Yes — latency sits under one second, and on τ-Voice it responds about 40% faster than GPT-Live-1 (Astra).
What does Grok Voice 3.0 cost to run?
Nothing changed: the Voice Agent API bills $0.08 per minute, or $4.80 per hour — identical to Grok Voice 2.0.
Does it work for live customer support?
It does. Starlink routes 20,000 calls a day through it in several languages, and around seven in ten finish without a human taking over.
Will existing grok-voice-latest agents break?
No. Agents already on grok-voice-latest were migrated in the background, so there is nothing for you to update.
Ship Your Voice Agent with Grok Voice 3.0
Give your product the top-ranked speech-to-speech model: sub-second replies, enterprise-grade call resolution, and flat $0.08/min Voice Agent API pricing.