# Websocket API for OpenAI

**URL:** <https://community.livekit.io/t/websocket-api-for-openai/374>\
**Category:** Getting Started\
**Created:** [February 24, 2026, 12:30pm UTC](https://community.livekit.io/t/websocket-api-for-openai/374 "2026-02-24T12:30:29Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Enrique\_Alcazar](https://sea1.discourse-cdn.com/flex001/user_avatar/community.livekit.io/enrique_alcazar/32/472_2.png) [@Enrique\_Alcazar](https://community.livekit.io/u/Enrique_Alcazar)\
**Post date:** [February 24, 2026, 12:30pm UTC](https://community.livekit.io/t/websocket-api-for-openai/374/1 "2026-02-24T12:30:29Z")

</div>

OpenAI recently released a websocket API (For their regular models not only realtime)

This should reduce latency since we don’t have to establish a new connection per generation.

Are there plans to adapt this? I imagine is going to require significant redesign of some core sections of the livekit openai plugin.

Is this a good first PR to work on?

---

<div class="post-metadata">

**Author:** ![CWilson](https://sea1.discourse-cdn.com/flex001/user_avatar/community.livekit.io/cwilson/32/135_2.png) [@CWilson](https://community.livekit.io/u/CWilson)\
**Post date:** [February 24, 2026, 1:25pm UTC](https://community.livekit.io/t/websocket-api-for-openai/374/2 "2026-02-24T13:25:37Z")

</div>

I saw that. Looks pretty fast too. We are looking into it. That would have been a great PR 🙂 But I believe the team already has one in flight.