Skip to content
View original post on X: LMSYS OrgOfficial
Published
32/100

SGLang-Omni adds full-duplex real-time interaction across three model types

AILMSYS Org says SGLang-Omni now supports real-time interaction through a unified runtime covering native full-duplex models, turn-based multimodal models, and cascaded systems. In single-session tests on one H200, the Qwen3.8-27B cascade with thinking off averaged 0.17s stop latency and 1.32s response latency.
LMSYS OrgVerified on XOfficial
@lmsysorg
View on X

🚀 Full-Duplex Interaction with SGLang-Omni

We've expanded SGLang-Omni's support for real-time interaction with a unified runtime abstraction for:
- Native full-duplex models
- Turn-based multimodal models
- Cascaded systems

In single-session tests on 1× H200, the Qwen3.8-27B cascade (thinking off) averaged 0.17s stop latency and 1.32s response latency.

Try it out! Blog and demos below 👇 https://www.sglang.io/blog/omni-interactive-runtime

Source details

LMSYS Org · x.comPublished · added here

Original title: 🚀 Full-Duplex Interaction with SGLang-Omni

Topics

Voice & audioOpen-source ecosystemDeployment & engineeringProduct updates