# Tether AI Research Unveils 254 Million-Parameter Speech Model

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/31774
Published: 2026-10-14T12:12:44.000Z
Updated: 2026-10-14T12:12:44.000Z
Section: Technology

> QVAC TranscribePsy is designed to run offline on devices, keeping audio on the device instead of sending it to cloud servers.

Tether AI Research announced QVAC TranscribePsy, a 254 million-parameter speech recognition model designed to process audio on devices without sending it to cloud servers.

The model can handle audio from smart glasses connected through Bluetooth or Wi-Fi. It can also run on a humanoid robot’s onboard processor.

TranscribePsy is intended to give developers an offline option for speech recognition, with low-latency processing that keeps audio on the device.