# SGLang Omni offers one runtime for full-duplex voice AI models

> Original title: Voice AI that can listen while it talks.

- Source: SGLang
- Published: 2026-11-09T02:26:02.000Z
- Bad Influence: https://badinfluence.news/items/upu0z1fcuta18qyavthqvmlia
- Original: https://x.com/sgl_project/status/2119617478510968885

## Summary

SGLang Omni contributors Chenyang Zhao and Jintao Qu describe a runtime that serves full-duplex models, omni models, and ASR–LLM–TTS pipelines. The runtime handles interruptions and keeps sessions persistent.

## Post

Voice AI that can listen while it talks.

Chenyang Zhao and Jintao Qu (SGLang Omni) on one runtime for full-duplex models, omni models and ASR–LLM–TTS pipelines, with interruption handling and persistent sessions.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWpkAKUaMAAEUfQ.jpg&mode=thumb&exp=1794528000&sig=4a65915796c7ceea)