Tasks and inputs
Grok Voice 3.0 is listed for Audio-to-audio, Text → Speech, Speech → Text. It accepts audio, text and produces audio, text. Use the publisher’s input format and preprocessing for your chosen deployment.
Grok Voice 3.0 is xAI's speech-to-speech model for real-time voice agents, replacing Grok Voice 2.0 at the same $0.08-per-minute price.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: Indicative: $0.08 per minute ($4.80/hour) plus $0.004 per text input; unchanged from Grok Voice 2.0. Confirm in your xAI console.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Nothing to migrate, nothing to re-price. If your agent points at grok-voice-latest, it moved to Grok Voice 3.0 the day xAI shipped it, in October 2026, and the bill stayed at $0.08 per minute. The interesting question is not whether to switch; it is whether the new model changes what you can hand to a voice agent.
| Item | Value |
|---|---|
| Model ID | grok-voice-3.0 |
| Alias | grok-voice-latest now resolves here |
| Type | Speech-to-speech (Voice Agent API) |
| Price (indicative) | $0.08/min ($4.80/hour) + $0.004 per text input |
| Regions | us-east-1, eu-west-1, us-saltlake-2 |
| Replaces | grok-voice-2.0 (deprecated) |
The upgrade path is one line: set model to grok-voice-3.0 or the alias. The API surface is unchanged. Prices are indicative, so confirm them in your xAI console.
xAI reports first place on Artificial Analysis' speech-to-speech index at 86.9, up from 81.3 for the previous version. In the same comparison, Gemini 3.8 Live in its Extended Thinking configuration sits at 82.6. Other publisher-reported numbers:
xAI says latency is under one second and that the model already handles millions of daily conversations across Tesla and SpaceX, including Starlink support calls where one agent verifies customers, troubleshoots hardware and issues credits. Those are deployment claims from the vendor, not independent measurements.
xAI's model listing tags the modality as text-to-text even while describing the model as speech-to-speech. It is a labelling slip; the endpoint is a voice endpoint. The exact date is also loose: xAI's notes say only "October", and the 14 October date comes from launch-day reporting. Neither affects integration, but both matter if you track releases.
Benchmarks rarely capture your callers' accents, your hold music or your CRM's slow lookups. Record twenty real calls (with consent), script them, and play them against the old and new IDs while 2.0 still answers. Measure three things: how often a caller has to repeat themselves, how the agent handles an interruption mid-sentence, and how many calls finish without escalation. Because the per-minute rate is the same, any reduction in call length is pure saving.
Pair it with Grok Voice Transcribe 2.0 if you need clean transcripts for QA, and see the rest of xAI's catalogue on the xAI provider page.
Compare the model on the work you need to complete.
Grok Voice 3.0 is listed for Audio-to-audio, Text → Speech, Speech → Text. It accepts audio, text and produces audio, text. Use the publisher’s input format and preprocessing for your chosen deployment.
Compare pronunciation, voice consistency and time to the first playable audio. Check the service’s billing unit for the actual script.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Alibaba
Qwen-Audio-3.1-Realtime Flash is Alibaba's lighter full-duplex speech model with audio and text in and out, function calling, web search and a 262K context.
ElevenLabs
Eleven Realtime v1 Max is ElevenLabs' reasoning speech-to-speech agent for support and multi-step tool workflows, OpenAI Realtime-compatible.
ElevenLabs
Eleven Realtime v1 Mini is ElevenLabs' fast, low-cost speech-to-speech voice agent stack, with an OpenAI Realtime-compatible API and 90+ languages.
Amazon
Nova 2.5 Sonic is Amazon's speech-to-speech model on Bedrock for real-time voice agents, with async tool calling, polyglot voices and a 256K context.
Alibaba
Qwen-Audio-3.1-Realtime-Plus supports duplex speech conversation, voice cloning, tools and a 262K-token context through its realtime interface.
Gemini 3.8 Live is Google's model for interactive audio conversations with text, image, audio and video input. Its Live API response modality is audio; text is available through transcription.