xAI · 2026-10-14 · major
Grok Voice 3.0 — xAI's voice model tops the speech-to-speech index
Grok Voice 3.0 is the new speech-to-speech model in xAI's Voice Agent API. It scores 86.9 on the Artificial Analysis Speech-to-Speech Index, first place, and keeps the $0.08 per minute price of Grok Voice 2.0.
xAI's third-generation voice model ranks first on Artificial Analysis' speech-to-speech index at the same price.
Key specs
| Speech to speech index | 86.9 |
|---|---|
| Full duplex bench | 97.0% |
Quick facts
| Maker | SpaceXAI (formerly xAI) |
|---|---|
| Model id | grok-voice-3.0 (alias grok-voice-latest) |
| Price | $0.08 / minute of audio ($4.80 / hour) |
| Availability | Voice Agent API, released 14 Oct 2026 |
| Replaces | grok-voice-2.0 (now deprecated) |
Benchmarks
| Grok Voice 3.0 | 86.9 pts | |
|---|---|---|
| GPT-Live-1 (Astra, Medium) | 83.2 pts | |
| Gemini 3.8 Live (High) | 82.6 pts | |
| Grok Voice 2.0 | 81.3 pts |
| GPT-Live-1 (Astra, Medium) | 74.5% | |
|---|---|---|
| Grok Voice 3.0 | 71.6% | |
| Gemini 3.8 Live (High) | 68.6% | |
| Grok Voice 2.0 | 56.5% |
Pricing
| Voice Agent API · $4.80 per hour, same as Grok Voice 2.0 | $0.08 / minute of audio |
|---|
What is it?
Grok Voice 3.0 is the speech-to-speech model behind the Voice Agent API from SpaceXAI, the company formerly called xAI. It listens and talks back in real time and is built for support and sales calls. It was released on 14 October 2026.
How does it work?
The API does not change. The alias grok-voice-latest now points to grok-voice-3.0, so agents using the alias moved to the new model automatically, and grok-voice-2.0 is deprecated. On Artificial Analysis tests the model leads on turn-taking and interruptions (97.0% on Full Duplex Bench) and speech reasoning (98.5% on Big Bench Audio).
Why does it matter?
Voice agents fail when they talk over callers or cannot finish a task, and Grok Voice 3.0 scores best on both while costing $0.025 per Big Bench Audio task against $0.030 for GPT-Live-1. SpaceXAI says Starlink's Grok Voice agent answers about 20,000 calls a day and resolves roughly 70% without a human.
Who is it for?
teams building phone and in-app voice agents
Frequently asked questions
- How much does Grok Voice 3.0 cost?
- Grok Voice 3.0 costs $0.08 per minute of audio, or $4.80 per hour, in xAI's Voice Agent API. That is the same price as Grok Voice 2.0. Artificial Analysis puts its cost per Big Bench Audio task at $0.025, against $0.030 for GPT-Live-1 and $0.055 for GPT-Realtime-2.1 (High).
- Do I need to change my code to use Grok Voice 3.0?
- No code change is needed if your agent uses the grok-voice-latest alias: xAI's release notes say the alias now resolves to grok-voice-3.0 and sessions were upgraded automatically. Agents pinned to grok-voice-2.0 should switch, because xAI marks that model id as deprecated. The API and price stay the same.
- Is Grok Voice 3.0 better than GPT-Live-1?
- On Artificial Analysis' figures, Grok Voice 3.0 beats OpenAI's GPT-Live-1 on the Speech-to-Speech Index (86.9 vs 83.2), Full Duplex Bench (97.0% vs 94.9%) and Big Bench Audio (98.5% vs 90.1%). GPT-Live-1 still wins the τ-Voice support-call test, 74.5% to Grok Voice 3.0's 71.6%.
- Where is Grok Voice 3.0 used today?
- SpaceXAI says Grok Voice 3.0 runs in Tesla vehicles for hands-free requests such as checking a calendar or ordering coffee, and powers Grok Bot's voice mode. Starlink call centers use a Grok Voice agent that answers about 20,000 calls a day in several languages and resolves roughly 70% without a human.
Try it
wss://api.x.ai/v1/realtime?model=grok-voice-3.0