xAI has launched the Grok Voice Agent API, enabling developers to create voice agents that speak dozens of languages, access tools, and search real-time data. The API uses the same technology as Grok Voice in apps and Tesla vehicles, providing a proven, versatile solution for developers.

Built entirely in-house, the Grok Voice stack includes custom voice activity detection, tokenisation, and audio models. Fine-grained control allows rapid iteration, making Grok one of the fastest and most intelligent voice agents on the market.

Benchmark tests show Grok achieves first audio output in under one second, nearly five times faster than competitors.

Grok can automatically detect and respond in the user's language, with native pronunciation and seamless mid-conversation language switching. Expressive voices like Ara, Eve, and Leo enable natural interaction and handle specialised terminology in healthcare, finance, and legal sectors.

The API integrates seamlessly with Tesla, enabling tasks like route planning and vehicle status queries, and supports custom tools or real-time searches across X and the web. Developers can test voices in the browser playground, with upcoming standalone text-to-speech and speech-to-text features.