Gemini 4 Argon
Analyze multimodal inputs and generate text for complex coding, scientific reasoning, and tool-connected agent applications.
Grok Voice 3.0
docs.x.ai
Build real-time conversational voice applications with streaming speech, configurable sessions, tools, and SIP phone integration.
Visit website
Pricing
Paid
Available on
API, Browser integration, Mobile integration, SIP telephony
Category
Ai Audio
What is Grok Voice 3.0?
Grok Voice 3.0 is a speech-to-speech model for developers building voice assistants, phone agents, and interactive audio applications. Its API streams audio and text bidirectionally over WebSocket.
The versioned identifier is grok-voice-3.0. Applications using grok-voice-latest automatically receive this release. Voice 3.0 retains Voice 2.0’s API and pricing; Voice 2.0 is deprecated.
Key features
Bidirectional conversation streaming
Send audio or text and receive streamed audio and text responses through realtime session events.
Configurable voice sessions
Set instructions, built-in or custom voices, reasoning effort, and automatic voice activity detection or manual text turns.
Tools within conversations
Configure function calling, file search, web search, X search, and MCP tools.
Browser and mobile authentication
A server issues short-lived ephemeral tokens so client applications can authenticate without receiving the underlying API key.
SIP call routing
Route PSTN, contact-center, or PBX calls into voice sessions using registered phone numbers and incoming-call webhooks.
Who is it for?
Developers adding an in-app assistant
Build a spoken interface that lets users converse with an application.
Support teams developing phone agents
Connect an existing telephone workflow to an agent with configured instructions and functions.
Product teams connecting voice to information
Create a conversational interface that consults configured search tools during a session.
How to use Grok Voice 3.0
- Create an API key through the developer console and select grok-voice-3.0 to pin the release.
- For browser or mobile clients, configure a server endpoint to issue ephemeral tokens.
- Connect to the realtime WebSocket endpoint with the appropriate authentication.
- Configure instructions, voice, turn detection, and tools through a session update.
- Send audio or text and handle response events.
Grok Voice 3.0 pricing
Paid
Voice API access is metered by conversation minutes and text inputs. Provider-operated tools have separate invocation charges, so tool-enabled sessions can cost more than the base voice rate.
Grok Voice 3.0 Speech to Speech
USD 0.08 per minute, plus USD 0.004 per text input
Metered API usage per minute and per text input.
The voice rate equals USD 4.80 per hour. Separate tool invocation charges may apply.
What to know before you use it
- Browser and mobile integrations require server-side token issuance. The underlying API key must stay out of client-side code.
- SIP deployment requires carrier or PBX routing configuration and incoming-call webhook handling. Provisioning provider-owned phone numbers through the API is unsupported.
- Conversation resumption requires opt-in when caching and reconnecting. Cached history expires after 30 minutes of inactivity.
Frequently asked questions
Will the latest-model alias upgrade my application automatically?
Yes. grok-voice-latest points to Grok Voice 3.0. Select grok-voice-3.0 to pin this release.
Can it answer calls through an existing phone system?
SIP integration routes carrier, contact-center, or PBX calls into speech-to-speech sessions after number registration, webhook configuration, and routing setup.
Do browser users need the application’s API key?
No. Your server obtains an ephemeral token and passes it to the browser for authentication.