Grok Voice 3.0
A practical guide to the speech model, its API and what a conversation costs.
Understand the release, compare the published results, and estimate your audio budget before building a voice agent.
By AIVoiceChanger · Checked · Independent guide; no xAI affiliation.
What is the Grok Voice 3.0 speech model?
Grok Voice 3.0 is xAI’s speech-to-speech model for live voice agents, announced on October 14, 2026. The launch announcement reports sub-second latency and a first-place score of 86.9 on the Artificial Analysis Speech-to-Speech Index. These are published results, not measurements from this site. Read the announcement ↗
Choose the workflow before the model: a support assistant needs to understand a caller and respond; a creator changing a recorded narrator needs voice transformation. Those tasks require different controls and different quality checks.
Grok Voice 3.0 vs 2.0: what changed?
| Measure | Grok Voice 3.0 | Grok Voice 2.0 |
|---|---|---|
| Speech-to-Speech Index | 86.9 | 81.3 |
| Task success rate | 94.8% | 94.6% |
| Full Duplex Bench | 97.0% | 95.1% |
In the same launch table, GPT-Live-1 (Astra, Medium) scores 83.2 and Gemini 3.8 Live (Extended Thinking, High) scores 82.6 on the overall index. Source and benchmark context ↗
The task-success difference is small relative to the overall-index gap. Test your own call scenarios before treating an aggregate ranking as a reason to switch.
The API documentation says 2.0 is deprecated, the endpoint and events stay the same, and grok-voice-latest now resolves to 3.0. Pin grok-voice-3.0 when you want a fixed version. Migration documentation ↗
Grok Voice 3.0 pricing calculator
The published API price is $0.08 per audio minute, plus $0.004 per text input. At that audio rate, 10 minutes is $0.80, one hour is $4.80 and 1,000 minutes is $80. This estimate is separate from AIVoiceChanger subscriptions.
Audio estimate only: minutes × $0.08. Text inputs ($0.004 each), tools, telephony and other applicable charges are excluded. No API request is made.
For a pilot, estimate total conversation minutes across all calls. Then compare the estimate with actual console usage before committing to a larger rollout.
How to get started with the Grok Voice 3.0 API
- Create access in the official API console and configure your server credentials.
- Connect your server to the versioned WebSocket endpoint below.
- Configure the voice and instructions with
session.update; send conversation input and handle response events. - For browser clients, use short-lived ephemeral tokens issued by your backend. Keep the permanent key server-side.
wss://api.x.ai/v1/realtime?model=grok-voice-3.0This endpoint is a connection reference, not a complete audio client. Use the official quickstart and tester apps for capture, streaming and playback. We have not executed a paid API session for this guide.
How to evaluate it for your use case
Use the same scripts and recording conditions for each candidate. Keep a short scorecard instead of relying on an impressive single demo.
Conversation quality
Interrupt an answer, correct a name, pause mid-sentence and ask a follow-up. Record whether the agent recovers without losing your intent.
Your real audio
Include quiet rooms, background noise and the accents your audience uses. Compare time to first audible response and intelligibility.
Task completion
Define a successful outcome in advance. Count wrong actions and failed handoffs as failures even when the voice sounds convincing.
Cost per outcome
Track the complete bill and the number of successful calls. Longer conversations or retries can erase a low per-minute price.
Change a voice in our online studio
Upload audio, record a take or enter text to explore our voice presets. The studio uses ElevenLabs; it is a separate voice-changing tool, not a Grok demo.
Open the voice changer →Grok Voice 3.0 FAQ
What is the Grok Voice 3.0 speech model?
It is a real-time speech-to-speech model for conversational agents. It takes part in a conversation; an audio voice changer instead transforms an existing performance into a different voice.
Is Grok Voice 3.0 free?
The API is metered. The published audio rate is $0.08 per minute, with text inputs and other applicable charges separate. This page’s cost calculator is free and does not call the API.
Can I try Grok Voice 3.0 here?
Use the official console linked above to explore Grok. AIVoiceChanger is an independent site and has not integrated Grok Voice 3.0. Our separate voice studio uses ElevenLabs.
Is this a hands-on Grok Voice 3.0 review?
No. This is a source-based guide checked on October 17, 2026. Benchmark figures are attributed to the launch announcement; we have not run a paid API test or an independent listening study.
Sources & update notes
Checked October 17, 2026. Pricing and model aliases can change. This guide summarizes documentation and offers an evaluation checklist; it does not claim independent benchmark testing.