# Grok Voice 3.0

SpaceXAI, released 2026-10-14

**Verdict: conditional.** Grok Voice 3.0 is SpaceXAI's speech-to-speech model for real-time voice agents, priced at $0.08 per minute, the same as Grok Voice 2.0. Artificial Analysis shows the highest Speech-to-Speech Index score, 86.9, with 0.81 seconds to first audio, but xAI's other benchmark and customer figures are its own. Conditional because we have not tested it and the grok-voice-latest alias moved to it automatically. It suits teams building phone or in-app voice agents on a managed model. Last verified 2026-10-18.

## Pricing

- Input: $0.08 / min of audio (speech-to-speech)
- Output: Included in per-minute audio rate
- Context: Voice Agent API; $4.80 / hr

## Strengths

- Real-time voice agents over WebSockets: xAI's voice docs describe speech-to-speech agents on wss://api.x.ai/v1/realtime with server-side voice activity detection and tool use such as web search
- Low time to first audio: Artificial Analysis lists 0.81 seconds for Grok Voice 3.0, and xAI claims class-leading latency under one second
- Callers who interrupt: xAI reports 97% on Full Duplex Bench, a measure of pauses, turn-taking and interruptions (xAI's figure)
- Per-minute budgeting: $0.08 per minute ($4.80 per hour) on the pricing page, plus a separate text-input line (unit unresolved)
- Upgrading without code changes: xAI's release notes say sessions on the grok-voice-latest alias are routed to the new model and that the API is unchanged
- Teams that need pinned behaviour: grok-voice-latest was switched over automatically, so pin grok-voice-3.0 explicitly if the model must not change under you
- Buyers who need independent customer evidence: the Tesla, SpaceX and Starlink usage figures, and the EVA-Bench and tau-Voice results, are xAI's own claims
- Exact cost modelling from the page we read: the pricing page also lists a separate text-input line whose unit we could not resolve, so check it before budgeting
- Staying on Grok Voice 2.0: the pricing page marks grok-voice-2.0 as deprecated and xAI tells users to migrate to grok-voice-3.0 or grok-voice-latest

## Benchmarks

- Artificial Analysis Speech-to-Speech Index: 86.9 (https://artificialanalysis.ai/speech-to-speech)
- Full Duplex Bench (xAI-reported): 97% (https://x.ai/news/grok-voice-3)
- tau-Voice (xAI-reported): 71.6% (https://x.ai/news/grok-voice-3)
- EVA-Bench (xAI-reported): 74% (https://x.ai/news/grok-voice-3)

## Sources

- [xAI docs pricing (STT $0.10/hr batch, $0.20/hr streaming)](https://docs.x.ai/developers/pricing)
- [xAI docs: release notes (grok-voice-latest routed to Grok Voice 3.0)](https://docs.x.ai/developers/release-notes)
- [xAI docs: Voice Agent API guide](https://docs.x.ai/docs/guides/voice)
- [xAI: Grok Voice 3.0 announcement (October 14, 2026)](https://x.ai/news/grok-voice-3)
- [Artificial Analysis: Speech-to-Speech leaderboard (Grok Voice 3.0 86.9, 0.81 s)](https://artificialanalysis.ai/speech-to-speech)

Canonical: https://neomanex.com/models/grok-voice-3-0