Maintaining an existing 1.0 deployment
Teams can pin the versioned model while they evaluate whether Think Fast 2.0 changes prompts, tools, voices, latency, or customer outcomes.
Independent tool overview
Grok Voice Think Fast 1.0 at a glance
Grok Voice Think Fast 1.0 is xAI's previous-generation real-time speech-to-speech model for phone support, sales, scheduling, and other tool-using voice agents; it remains documented but is deprecated in favor of Think Fast 2.0.
Visit the official Grok Voice Think Fast 1.0 site ↗
- Product type
- Real-time speech-to-speech API model
- Current status
- Available but deprecated
- Recommended successor
- Grok Voice Think Fast 2.0
- 1.0 API price
- $0.05 per audio minute
- Connection options
- WebSocket, SIP, and supported integrations
- Best suited to
- Existing integrations that require the pinned 1.0 model
- Last reviewed
- August 29, 2026
Overview
What Grok Voice Think Fast 1.0 is
Grok Voice Think Fast 1.0 was introduced as xAI's flagship API model for real-time voice agents. It was designed to listen and speak in a single live session while reasoning through ambiguous requests, calling business tools, collecting structured details, and handling interruptions, accents, noise, and multi-turn conversations.
The model is still listed in xAI's speech-to-speech documentation as a versioned option, but xAI now marks it deprecated. On August 5, 2026, the grok-voice-latest alias moved to Grok Voice Think Fast 2.0. New deployments should use the latest alias or pin 2.0 after testing; pinning 1.0 is mainly useful when an existing production workflow depends on its exact behavior.
Think Fast 1.0 can be connected over xAI's real-time API and used with custom functions plus xAI tools such as file search, web search, X search, and remote MCP servers. The surrounding Grok Voice platform also supports WebSocket and SIP integrations, configurable voices, turn detection, live transcripts, and human handoff patterns.
This is infrastructure for developers and operations teams, not a finished call-center replacement. A production deployment still needs telephony, identity checks, permission boundaries, reliable backend tools, compliance disclosures, escalation paths, monitoring, and careful testing with real accents, noise, interruptions, and failure cases.
Use cases
Who Grok Voice Think Fast 1.0 is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Tool-using phone support
The model was built for multi-step conversations that retrieve account information, follow policy, update records, and escalate to a person when required.
Sales and scheduling workflows
Voice agents can qualify callers, check availability, book appointments, place structured tool calls, and confirm important details aloud.
Multilingual real-time conversations
xAI positioned the model for noisy, accented, and multilingual call environments where interruptions and corrections are normal.
Capabilities
Core Grok Voice Think Fast 1.0 features
Native speech-to-speech sessions
The model accepts live audio and returns streamed speech in the same real-time conversation rather than requiring teams to manually chain separate transcription, text-model, and speech APIs.
Background reasoning
xAI designed Think Fast 1.0 to reason while the conversation continues, supporting complex and ambiguous requests without adding a separate visible reasoning step.
Function and tool calling
Agents can invoke developer-defined functions and supported xAI tools during a call, then use the results to continue the conversation.
Structured data collection
The launch focused on capturing and reading back names, addresses, phone numbers, account identifiers, and corrected details during natural speech.
Interruptions and turn detection
Server-side voice activity detection can identify turns and allow callers to interrupt, with configurable silence, threshold, and padding behavior.
Telephony support
The broader API supports SIP and common telephony audio codecs for routing phone-system and contact-center calls into real-time sessions.
Multiple voices and languages
Developers can choose supported built-in or custom voices, while the voice system is designed for multilingual speech and code-switching.
Session instructions and controls
Teams can configure the system prompt, tools, reasoning behavior, voice, audio formats, turn detection, and other session parameters.
Version pinning
The explicit grok-voice-think-fast-1.0 identifier keeps an integration on this model instead of following the grok-voice-latest alias to newer releases.
Process
How the Grok Voice Think Fast 1.0 workflow works
Step 1
Decide whether 1.0 is still necessary
For a new agent, start with Think Fast 2.0. Keep 1.0 only when testing shows that an existing production workflow needs its current behavior during migration.
Step 2
Map the conversation and escalation rules
Define the permitted tasks, required disclosures, identity checks, prohibited actions, failure handling, and exact points where a human takes over.
Step 3
Connect a real-time session
Authenticate server-side or use short-lived client tokens, then connect over the supported real-time API with the versioned model name.
Step 4
Add narrowly scoped tools
Expose only the functions the agent needs, validate every argument, enforce authorization in the backend, and make consequential writes reversible where possible.
Step 5
Configure voice and turn behavior
Choose the voice, input and output formats, interruption behavior, silence timing, and any pronunciation replacements needed for the use case.
Step 6
Test realistic failure cases
Use noisy telephony audio, accents, interruptions, corrections, ambiguous requests, tool failures, prompt injection attempts, and requests that must be refused or escalated.
Step 7
Compare against 2.0 and migrate
Run matched evaluations for task completion, transcription, latency, tool accuracy, cost, and customer outcomes, then move to 2.0 unless 1.0 has a documented business requirement.
Cost
Grok Voice Think Fast 1.0 pricing and free plan
xAI currently lists the deprecated Think Fast 1.0 model at $0.05 per audio minute, or $3.00 per audio hour, plus $0.004 per text input. Think Fast 2.0 costs more per audio minute and is the recommended current model.
Grok Voice Think Fast 1.0
$0.05 per audio minute
Pay-as-you-go access to the previous-generation, deprecated speech-to-speech model.
- $3.00 per audio hour
- $0.004 per text input
- Telephony and other external service costs are separate
- Model is still documented but deprecated
Grok Voice Think Fast 2.0
$0.08 per audio minute
The current flagship speech-to-speech model and recommended upgrade for new and existing integrations.
- $4.80 per audio hour
- $0.004 per text input
- The grok-voice-latest alias points to 2.0
- Test migration behavior before changing a production call flow
Pricing checked . Check current pricing at the source ↗
Assessment
Grok Voice Think Fast 1.0 strengths and limitations
Where it stands out
- Combines listening, reasoning, tool use, and spoken responses in one real-time API
- Designed for noisy, interrupted, accented, and multi-step phone conversations
- Supports custom functions and multiple xAI-provided tool types
- Version pinning gives existing deployments time to test a controlled migration
- Lower listed audio-minute price than its 2.0 successor
- Works with SIP and real-time WebSocket integration patterns
- Supports configurable voices, turn detection, audio formats, and session instructions
What to consider
- xAI marks Think Fast 1.0 deprecated and recommends the newer Think Fast 2.0 model
- The grok-voice-latest alias no longer points to 1.0, so applications that require it must use the explicit versioned name
- A lower per-minute price may not offset weaker transcription, conversational, latency, or tool-use performance compared with the successor
- The API is a model layer, not a complete support operation; teams still need telephony, business integrations, observability, escalation, and quality assurance
- Voice agents can misunderstand callers or perform the wrong tool action, especially under ambiguous or adversarial conditions
- High-stakes actions require server-side authorization and human approval instead of relying only on the model prompt
- Calls may contain personal, financial, health, or account data that requires appropriate consent, retention, access, and redaction controls
- Per-minute model pricing excludes phone numbers, carrier charges, engineering, monitoring, storage, and third-party tools
- Languages, voices, codecs, events, and compatibility details should be verified against the current API documentation before implementation
- Keeping a deprecated model increases future migration risk and may delay access to quality and reliability improvements
Compare
Grok Voice Think Fast 1.0 alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Sales
Vapi
A developer platform for assembling and operating voice agents across model, speech, telephony, and tool providers.
Explore Vapi →Business Operations
Retell AI
A managed conversational voice-agent platform focused on production phone calls and business workflows.
Explore Retell AI →Sales
Bland AI
A phone-agent platform aimed at automating customer calls with configurable workflows and integrations.
Explore Bland AI →Questions
Grok Voice Think Fast 1.0 FAQs
Is Grok Voice Think Fast 1.0 still available?
xAI's current documentation still lists the versioned grok-voice-think-fast-1.0 model, but marks it deprecated. Teams should confirm availability for their account and plan a migration.
What replaced Grok Voice Think Fast 1.0?
Grok Voice Think Fast 2.0 is the current flagship successor. The grok-voice-latest alias moved from 1.0 to 2.0 on August 5, 2026.
How much does Grok Voice Think Fast 1.0 cost?
xAI currently lists 1.0 at $0.05 per audio minute, equal to $3.00 per audio hour, plus $0.004 per text input. Telephony and other infrastructure are separate.
Should a new project use Think Fast 1.0?
Usually no. xAI recommends the newer 2.0 model, and its latest alias points there. The versioned 1.0 model is most relevant to existing integrations that need a controlled migration.
Why would a team pin the 1.0 model?
Pinning prevents an automatic model change while a team compares prompts, tools, voices, latency, costs, and customer outcomes. It should be a temporary compatibility decision with a migration plan.
Can the model call business tools during a conversation?
Yes. The real-time API supports custom functions and tool types such as file search, web search, X search, MCP, and other configured functions. Backend systems must still enforce permissions and validate every action.
Does xAI provide a no-code voice-agent option?
Yes. xAI also offers a Voice Agent Builder in beta with telephony, knowledge retrieval, tools, guardrails, and observability. That product is separate from directly integrating the model API.
Can Grok Voice handle phone calls?
The speech-to-speech platform supports SIP and telephony-oriented codecs, allowing phone-system calls to be routed into real-time sessions. Carrier, number, routing, and compliance requirements still need to be handled.
Is a voice agent safe for high-stakes support actions?
Not by default. Teams should require identity verification, narrow tool permissions, server-side policy checks, auditable logs, human approval for consequential actions, and immediate escalation when confidence or authorization is insufficient.
Bottom line
Our Grok Voice Think Fast 1.0 verdict
Grok Voice Think Fast 1.0 was an important step toward fast, reasoning, tool-using voice agents, but it is now a compatibility model rather than the best default. Existing users can pin it while testing migration, but new deployments should start with Think Fast 2.0 and adopt the operational controls required for any agent that can speak to customers and change business records.
Visit Grok Voice Think Fast 1.0 website ↗Similar Agents Tools
View all Agents AI tools →
Gemini Enterprise Agent Platform
Gemini Enterprise Agent Platform - Google's successor to Vertex AI and platform for building, scaling, and governing enterprise agents across 200+ models
Workflows
Workflows - Mistral's enterprise tool for chaining AI agents into multi-step business processes
Workspace Agents
Workspace Agents - OpenAI's new Codex-powered team agents for shared multi-step workflows in ChatGPT and Slack
Manus Cloud Computer
Manus Cloud computer - No-code environment for deploying always-on agents, scrapers, and self-hosted apps
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.