Grok Voice 3.0: xAI’s Speech-to-Speech Model
Grok Voice 3.0 is xAI’s real-time speech-to-speech model for voice agents, announced October 14, 2026. Official positioning: class-leading latency under one second, $0.08 per minute (same price as Grok Voice 2.0), and the top score on Artificial Analysis’s Speech-to-Speech Index (86.9, vendor-cited). The API model id is grok-voice-3.0; grok-voice-latest already points at 3.0. This Toolsiftly page summarises xAI’s own announcement and docs only — it is not an independent lab review.
Grok Voice 3.0 at a glance
| Grok Voice 3.0 | |
|---|---|
| Maker | xAI |
| Announced | October 14, 2026 (x.ai news) |
| Type | Real-time speech-to-speech / voice-agent model |
| API model id | grok-voice-3.0 (see docs) |
| Alias | grok-voice-latest switched to 3.0 automatically per the announcement |
| Price | $0.08 / min ($4.80 / hr) — same as Grok Voice 2.0 |
| Headline metrics (vendor-cited) | AA Speech-to-Speech Index 86.9; Full Duplex Bench 97%; EVA-Bench accuracy 74% (ServiceNow quote on xAI news) |
What xAI says it is for
Per the announcement, Grok Voice 3.0 targets voice agents that need tool use, sub-second turns, and full-duplex conversation (pauses, interruptions). xAI states it already powers high-volume Tesla in-car conversations and Starlink call-centre traffic; those deployment claims are xAI’s, not independently verified here.
Benchmark table (from xAI’s announcement)
Figures below are copied from xAI’s Oct 14, 2026 post, which cites Artificial Analysis (October 13, 2026). Treat them as vendor-reported comparisons, not Toolsiftly measurements.
| Benchmark | Grok Voice 3.0 | Grok Voice 2.0 | GPT-Live-1 (Astra, Medium) | Gemini 3.8 Live (Ext. Thinking, High) |
|---|---|---|---|---|
| Overall AA Speech-to-Speech Index | 86.9 | 81.3 | 83.2 | 82.6 |
| Speech Reasoning (Big Bench Audio) | 98.5% | 97.2% | 90.1% | 97.7% |
| Human Preference (Speech Agent Arena Elo) | 1070 | 1011 | 1048 | 990 |
| Task Completion (Task Success Rate) | 94.8% | 94.6% | 87.4% | 89.1% |
| Conversational Dynamics (Full Duplex Bench) | 97.0% | 95.1% | 94.9% | 91.9% |
Pricing and availability
- Voice Agent API: $0.08 per minute.
- Drop-in: agents already on
grok-voice-latestmoved to 3.0 with no code change (per xAI). - Docs model page: Grok Voice 3.0 lists the model name and regions (us-east-1, eu-west-1, us-saltlake-2 on the docs page we fetched).
For request wiring (WebSocket / session.update), use the current Speech to Speech docs on docs.x.ai — those pages change with the API; this guide does not paste stale SDK snippets.
Grok Voice 3.0 vs 2.0 (short)
| 3.0 | 2.0 | |
|---|---|---|
| Status | Current (grok-voice-latest) |
Predecessor (deprecated per release notes cited in search) |
| Price | $0.08/min | Same |
| AA Index (vendor table) | 86.9 | 81.3 |
What this page is (and isn’t)
- No star rating or “best voice AI” ranking from us.
- No unofficial endpoints, leaked keys, or third-party wrappers.
- No medical/legal/financial advice about deploying voice bots in regulated industries.
Frequently asked questions
What is Grok Voice 3.0?
xAI’s October 14, 2026 speech-to-speech model for real-time voice agents. Official model id: grok-voice-3.0.
How much does Grok Voice 3.0 cost?
xAI lists $0.08 per minute ($4.80 per hour) on the Voice Agent API — the same price as Grok Voice 2.0.
What is the Grok Voice 3.0 speech model API name?
Use grok-voice-3.0. The grok-voice-latest alias already points at 3.0 according to xAI’s announcement.
Is Grok Voice 3.0 better than 2.0?
xAI’s announcement cites higher Artificial Analysis Speech-to-Speech Index (86.9 vs 81.3) and Full Duplex Bench (97% vs 95.1%) at the same price. We have not re-run those benchmarks.
Is Grok voice mode free?
API usage is billed at the list price above. Consumer Grok apps may include voice under a subscription or free tier that changes over time — check xAI’s current product pages. This guide only covers the Voice Agent API facts xAI published for Grok Voice 3.0.
Where are the official docs?
Start at x.ai/news/grok-voice-3 for the launch post and docs.x.ai for the model page and speech-to-speech API reference.