SpaceXAI, the AI division formed after xAI’s integration into SpaceX, has added its Grok voice models to the Vercel AI Gateway. The move gives developers access to a full voice stack, including realtime voice, text-to-speech, and speech-to-text, through a single API layer that already handles routing, observability, and budget controls.
The announcement landed on June 29, 2026, with Vercel posting the update to its changelog and xAI officials amplifying it on X. The framing from both sides was direct: “state of the art voice APIs from SpaceXAI” are now live on the Gateway.
What’s actually in the box
Three models shipped with the integration. The realtime voice model is listed as xai/grok-voice-think-fast-1.0, designed for low-latency, bidirectional interactions. Then there’s xai/grok-tts for text-to-speech and xai/grok-stt for speech-to-text.
Bidirectional WebSocket streaming is baked in, which is the technical requirement for anything resembling a real phone call or live voice assistant.
