xAI Launches Grok Voice Think Fast 2.0 Speech-to-Speech AI Model
Elon Musk's AI startup xAI has announced Grok Voice Think Fast 2.0, its most advanced speech-to-speech AI model. The new model, priced at $0.08 per minute of audio, is scheduled to replace the 1.0 version on August 5. This update enhances real-time voice interactions by performing reasoning concurrently with speech, significantly reducing latency and token usage. It marks a major step forward in the competitive voice AI landscape, outperforming several dedicated transcription models. The model scored 82.9% on the Artificial Analysis Speech-to-Speech Quality Index and completes tool calls before the agent finishes its first sentence. Internal tests show its transcription accuracy is 1.5 to 2.0 times better than Deepgram Nova 3 and ElevenLabs Scribe v2.
## BACKGROUND
Speech-to-speech AI models process spoken input directly and generate spoken output, bypassing the latency of traditional cascading systems that convert speech to text, process it, and convert it back to speech. xAI, founded by Elon Musk, develops the Grok family of AI models to compete with rivals like OpenAI and Google.