Grok Voice 3.0 costs the same as the model it replaces: $0.08 a minute through xAI's Voice Agent API, or $4.80 an hour, plus $0.004 per text input. What changed is everything else. xAI's new realtime speech-to-speech model tops Artificial Analysis's Speech-to-Speech Index at 86.9, answers in under a second, and is already taking real calls.
If your agent calls grok-voice-latest, you're on it already; the alias moved to grok-voice-3.0, and grok-voice-2.0 is marked deprecated on the models and pricing page (checked October 15, 2026).
Benchmark Scores
xAI's announcement cites Artificial Analysis results dated October 13, 2026:
| Benchmark | Grok Voice 3.0 | Grok Voice 2.0 | GPT-Live-1 (Astra, Medium) | Gemini 3.8 Live (Ext. Thinking, High) |
|---|---|---|---|---|
| AA Speech-to-Speech Index | 86.9 | 81.3 | 83.2 | 82.6 |
| Big Bench Audio | 98.5% | 97.2% | 90.1% | 97.7% |
| Speech Agent Arena Elo | 1070 | 1011 | 1048 | 990 |
| Task success | 94.8% | 94.6% | 87.4% | 89.1% |
| Full Duplex Bench | 97.0% | 95.1% | 94.9% | 91.9% |
The index blends speech reasoning, τ-Voice, and preference and task success from blind calls. Full Duplex Bench is the one to watch for phone work: it scores how a model copes with pauses, turn-taking and being interrupted.
On τ-Voice, simulated support calls with accents, noise and interruptions, Grok Voice 3.0 gets 71.6%. xAI calls that comparable to GPT-Live-1 (Astra) at 40% lower latency. ServiceNow's EVA-Bench for enterprise voice agents gives it the top accuracy, 74%.
Production Use
It runs millions of conversations a day in Tesla cars and powers Grok's voice mode. Starlink's call centers put about 20,000 calls a day through Grok Voice in several languages; xAI says roughly 70% end without a human and one in four sales calls closes before hanging up.
xAI's other audio products are priced separately: speech-to-text at $0.10/hr over REST, text-to-speech at $15 per million characters.
AI Rank Data
AI Rank's voice board follows OpenRouter's speech ranking, which is mostly text-to-speech, so Grok Voice 3.0 isn't on it. xAI's TTS model, x-ai/grok-voice-tts-1.0, ranked #8 by requests for October 8–14, 2026, behind Gemini 3.8 Flash TTS and Kokoro 82M.
→ Voice ranking: see the current speech model usage ranking on AI Rank (requests, Oct 8–14, 2026 window).
Compare before you choose: pricing changes and rankings move daily — check the current model rankings on AI Rank before you commit.
Or 📱 get the AI Rank app to follow them on your phone.
Prepared with automated editorial assistance and checked against official sources and dated AI Rank data on October 15, 2026. We did not run our own benchmarks for this article.