xAI's voice conversation agent AI, 'Grok Voice Think Fast 1.0,' has been released.



xAI has announced its flagship voice model, the ' Grok Voice Think Fast 1.0 .' This model allows for voice-based interaction with users and is described as being useful for customer support and telephone sales.

Grok Voice Think Fast 1.0 | xAI

https://x.ai/news/grok-voice-think-fast-1




The Grok Voice Think Fast 1.0 is a model with multiple advantages, including top-level intelligence and reduced response lag, enabling natural responses even in complex, multi-part voice interactions.

You can hear the actual audio on the following website. Entering text will start a voice conversation.

xAI Console — Grok API & Developer Tools
https://console.x.ai/playground/voice/agent#session

I actually tried it. While it pronounces Japanese with almost accurate accent, there are some discrepancies between the output text and the audio (for example, where it says '1885').

What does the voice of xAI's voice conversation agent AI 'Grok Voice Think Fast 1.0' sound like? - YouTube


Grok Voice Think Fast 1.0 has achieved the top position in the 'τ-voice Bench,' a benchmark test that evaluates voice agents under realistic conditions such as noise, accent, and interruptions.



xAI stated, 'This model has been thoroughly tested under the most demanding real-world conditions, including phone calls, background noise, strong accents, and frequent interruptions. It natively supports more than 25 languages, making it ideal for global deployment.'

Furthermore, it has a high ability to understand the speaker's pronunciation, and can seamlessly collect addresses, phone numbers, names, account numbers, and other information even if the speaker speaks quickly or repeats themselves. It can also perform inference in the background, allowing it to process complex queries and workflows in real time without affecting response speed.



Grok Voice Think Fast 1.0 is already supporting Starlink's customer support and sales, with one in five inquiries leading to a product purchase being handled through conversations with AI, and 70% of customer support inquiries being resolved by AI.

in AI, Posted by log1p_kr