Grok Voice Agent API

2025-12-17 · Publisher : xAI · Domain : Artificial Intelligence

xAI launches the Grok Voice Agent API to build multilingual voice agents with tool calls and real-time data, at a flat rate of $0.05 per minute.

What changes

  • Dozens of languages, automatic detection and language switching mid-conversation.
  • Real-time tools, including web and X search.
  • Ara, Eve and Leo voices, with auditory cues.
  • Compatible with the OpenAI Realtime API specification.

### Specifications - Average latency under one second, about five times faster than the closest competitor according to xAI. - Access: xAI console, voice playground, LiveKit plugin. - The release notes date the general availability of the Speech to Speech API to December 16, 2025 (the blog post is dated the 17th). - Model identifier, context and knowledge cutoff: not disclosed. ### Pricing Flat rate of $0.05 per minute of audio. ### Benchmarks Scores published by the vendor: first on Big Bench Audio (voice agent reasoning, numerical score not recorded); human preference in blind evaluation against OpenAI's Realtime API on pronunciation, accent and prosody. ### Key takeaways xAI enters the voice agent market through an API with a single per-minute price.

Links

Tags : Audio, xAI, Grok, Voix, API