Introducing Grok Voice 3.0
xAI ● Covered by 4 sources
xAI says Grok Voice 3.0 is its smartest speech model yet. It’s already in Tesla and SpaceX, and the new version is cheaper and faster to answer.
Based on reporting by xAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
xAI has launched Grok Voice 3.0, a speech-to-speech model it says is now the most intelligent one on the market. The pitch is simple: better answers, quicker replies, and lower cost. The company is also shipping it today in the API at the same price as Grok Voice 2.0.
The headline numbers are strong. In Artificial Analysis’s testing, Grok Voice 3.0 tops the Speech-to-Speech Index at 86.9. It also posts 97% on Full Duplex Bench, which tries to capture how well a system handles pauses, interruptions, and the messy rhythm of real conversation. On another measure, Big Bench Audio, it reaches 98.5% on speech reasoning.
But the more interesting bit is the tradeoff it claims to have solved. Grok Voice 3.0 scores 71.6% on τ-Voice, which covers support calls with accents, background noise, and interruptions, and xAI says that is comparable with GPT-Live-1 (Astra). The difference, according to the company, is latency: 40% lower. It also says the model costs $0.025 per task on Artificial Analysis’s Big Bench Audio run, below GPT-Live-1 (Astra, Medium) at $0.030 and GPT-Realtime-2.1 (High) at $0.055.
xAI is also leaning hard on real-world usage, not just benchmarks. It says Grok Voice powers millions of conversations a day across Tesla and SpaceX. In Tesla vehicles, drivers can use it for tasks like checking a calendar or answering an email, while Grok Bot handles larger asks such as ordering coffee or booking dinner. Starlink call centers use it too, handling 20,000 customer calls a day across multiple languages, with about 70% resolved without a human and one in four sales calls ending in a purchase before hang-up.
ServiceNow’s EVA-Bench adds another data point: Grok Voice 3.0 gets the highest accuracy score there at 74%, and ServiceNow says it leads when accuracy and customer experience are scored together. For enterprises, that combination matters more than a flashy demo. For everyone else, it is another reminder that voice AI is moving from novelty to infrastructure, which is exactly where the real money lives.
My take — AI-written commentary, not fact-checked reporting
This is the kind of launch that makes benchmark junkies happy and call-center managers nervous, which is probably the right split. The interesting part isn’t the bragging; it’s that xAI is selling speed, cost, and “good enough” reliability as one package. That’s how voice AI stops being a demo and starts replacing actual work, one support queue at a time.
Read more about this at: xAI
Related stories
SpaceXAI Releases Grok Voice Transcribe 2.0: A Speech-to-Text API Claiming 2x Accuracy Over 1.0 at $0.10 per Hour
MarkTechPost · 3 weeks ago ·
8
Anthropic’s Watermarks, Grok 4.6 Surges, Qwen’s Open Weights, Better Corrections for Speech Recognition
The Batch ·
4
Google Releases Gemini 3.8 Live and 3.8 Live Extended Thinking for Production Grade Voice Agents
MarkTechPost · 1 month ago ·
8