xAI has released Grok Voice 3.0, a new speech-to-speech AI voice model built for customer support calls and hands-free assistants. The company says it answers in under a second, tops an independent voice ranking, and costs the same as the version it replaces. Here is what is new, how it compares, and why it matters if you build or talk to voice agents.
What Grok Voice 3.0 Does
Grok Voice 3.0 listens to speech and replies with speech, without turning everything into text first. That design helps it keep up with natural conversation, including pauses, interruptions, and people talking over each other.
xAI says the model is already handling millions of conversations a day across Tesla and SpaceX. It is available now through xAI’s Voice Agent API for developers.
Pricing and Availability
The price is $0.08 per minute, or $4.80 per hour, which matches Grok Voice 2.0. Developers whose agents point to the “grok-voice-latest” setting were switched to the new model automatically, with no code changes needed.
How the New Voice Model Compares
xAI points to independent testing by Artificial Analysis, a firm that benchmarks AI models. In the figures xAI published, Grok Voice 3.0 ranked first on the Speech-to-Speech Index with a score of 86.9.
| Model | Speech-to-Speech Index | Customer support test (tau-Voice) |
|---|---|---|
| Grok Voice 3.0 | 86.9 | 71.6% |
| GPT-Live-1 (Astra, Medium) | 83.2 | 74.5% |
| Gemini 3.8 Live (Extended Thinking, High) | 82.6 | 68.6% |
| Grok Voice 2.0 | 81.3 | 56.5% |
Notice that Grok Voice 3.0 does not win every test. On tau-Voice, which simulates support calls with accents, background noise, and interruptions, OpenAI’s GPT-Live-1 scored a bit higher. xAI says its model is comparable there while running at 40% lower latency.
Other Scores xAI Highlighted
- 97% on Full Duplex Bench, which measures how well a model handles pauses, turn-taking, and interruptions
- 74% accuracy on EVA-Bench, ServiceNow’s open benchmark for enterprise voice agents, the top score there
- About $0.025 average cost per task in Artificial Analysis’s Big Bench Audio run, versus $0.030 for GPT-Live-1 and $0.055 for GPT-Realtime-2.1
- A big jump over its own predecessor on support calls, from 56.5% to 71.6%
Keep in mind that these comparisons come from xAI’s announcement. Benchmarks are a helpful guide, but real-world results depend on how a company sets up its agent.
Grok Voice in Cars and Call Centers
The most interesting claims are about real use. In Tesla vehicles, drivers can use Grok to check a calendar, reply to an email, or hand off bigger jobs like ordering coffee or booking dinner, all without touching a screen.
xAI says Starlink call centers use Grok Voice to answer 20,000 calls a day in several languages. According to the company, roughly 70% of those calls are resolved without a human, and one in four sales calls ends in a purchase before the caller hangs up.
A single agent built on the model can verify a customer, troubleshoot hardware, issue credits, and ship replacement equipment. If those numbers hold up elsewhere, it shows how far AI support agents have come in a short time.
Fast replies used to come at the cost of smart replies. The race in voice AI now is to deliver both, at a price that makes phone support cheaper than a human queue.
What This Means for You
For most people, the effect will show up the next time you call a support line or talk to your car. Voice bots are getting quicker and better at handling interruptions, so calls may feel less robotic.
- If you run a business, compare cost per minute and real call resolution rates, not just benchmark rankings.
- Test any voice agent with real accents, noisy rooms, and impatient callers before you go live.
- Keep an easy path to a human for billing disputes and sensitive account issues.
- As a caller, never share passwords or full card numbers just because a voice sounds official.
That last point matters because better voice AI also helps scammers, so it is worth knowing how to guard against AI voice scams. If you want to compare tools, our voice generation section covers many options. Grok also sits inside xAI’s broader plans, including the recent X subscription bundle.
For a wider view of where things are heading, see our pieces on chatbots in support and the latest AI news.
Sources
This article draws on xAI’s official announcement, Introducing Grok Voice 3.0, and coverage from Unite.AI, which detailed the published benchmark tables.
Conclusion
Grok Voice 3.0 is xAI’s bid to lead the fast-growing market for AI phone and voice agents. It launches at the same $0.08 per minute price, ranks first on an independent speech-to-speech index, and answers in under a second, though rivals still edge it on some support tests. Its real-world track record at Tesla and Starlink is its strongest selling point. If you are weighing voice AI for your business, it is worth a test, but judge it on your own calls rather than on headline scores.
