Grok Voice Agent API

December 18, 2025

TL;DR

Grok Voice Agent API launched for developers.
Built on the same stack as Grok Voice in mobile apps and Tesla vehicles.
Features multilingual capabilities (dozens of languages) with native-level proficiency and seamless language switching.
Ranks #1 on Big Bench Audio benchmark for audio reasoning.
Offers an average time-to-first-audio of less than 1 second, significantly faster than competitors.
Industry-leading cost-efficiency at $0.05 per minute.
Consistently preferred over OpenAI Realtime API in human evaluations for pronunciation, accent, and prosody.
Designed with Tesla as a critical design partner, powering Grok in millions of vehicles.
Enables integration of custom tools and leverages xAI's real-time search across X and the web.
Offers multiple expressive voices (Ara, Eve, Leo) and auditory cues for enhanced realism.
Compatible with OpenAI Realtime API specification and available via xAI LiveKit Plugin.
Upcoming releases include standalone text-to-speech and speech-to-text endpoints, and improved audio models.

Continue reading
the original article