xAI Launches Grok Voice 3.0 via Developer API
xAI's new speech-to-speech model claims sub-second latency, supports over 25 languages and is already handling millions of Tesla voice dialogues daily.
xAI announced Grok Voice 3.0, a voice dialogue AI model, on October 14 and began offering it to developers through the Grok Voice API, according to watch.impress.co.jp.
Grok Voice 3.0 is a speech-to-speech model that takes voice input and responds with voice. xAI claims sub-one-second latency. The model supports natural dialogue features such as pauses, speaker turns and interruptions, and handles noisy and accented environments across more than 25 languages.
Pricing is 0.08 dollars per minute, or 4.80 dollars per hour, unchanged from the previous model. The Grok Voice API also supports integration with external services and tools, allowing users to search for information, check schedules and make reservations based on conversation content.
The model is already in use in Tesla's in-car AI, where it processes millions of voice dialogues per day. Starlink's call center uses Grok Voice to handle about 20,000 inquiries per day and resolves about 70 percent of them without human operators.
Grok Voice 3.0 is available through the Voice Agent API. Existing API users specifying grok-voice-latest are automatically switched to the new model.
Quick answers
How much does Grok Voice 3.0 cost?
It is priced at 0.08 dollars per minute, or 4.80 dollars per hour, unchanged from the previous model.
How many languages does Grok Voice 3.0 support?
It handles noisy and accented environments across more than 25 languages.
Where is Grok Voice 3.0 already being used?
Tesla's in-car AI processes millions of voice dialogues per day with the model, and Starlink's call center handles about 20,000 inquiries per day, resolving about 70 percent without human operators.