We're having issues running gpt-realtime-2 on LiveKit - anyone else experiencing this? For example we can't specify reasoning level, and the model seems to have weird outputs

Hi Alexander D, ⚠️ This channel will be deleted soon ⚠️ Please use our new community at https://community.livekit.io for technical discussions and to post technical questions. The Slack community is not going away, but we are migrating technical questions to our new forum. Please see full details here.

Yes, a few teams have been seeing inconsistent behavior with gpt-realtime-2 on LiveKit, especially around unsupported/partially supported parameters like reasoning controls and occasional unstable outputs in streaming sessions. In many cases, it’s more of an integration/runtime compatibility issue between the realtime model and orchestration layer than the model alone. I’d recommend verifying which parameters are actually exposed/supported in your current LiveKit pipeline and testing with a minimal config first to isolate whether the issue is model-side or session handling related. I can also help debug the setup and stabilize the realtime behavior if needed. @witty-dentist-65028