Hey, I saw that OpenAI released GPT-Realtime 2 yesterday. Has anyone tested it in production yet? Curious how it compares to GPT-Realtime and GPT-Realtime 1.5 in terms of latency and overall intelligence/reasoning quality. Any feedback would be appreciated, trying to figure out whether the switch is worth it.
I’ve not tried in productions, but it seems quite a bit better at giving results compared to the previous model.
@Petar_Slatinac Practical caveat for anyone testing: gpt-realtime-2 is a thinking model, so each turn carries reasoning latency, and livekit-plugins-openai doesn’t yet expose reasoning_effort to dial it down. Support gpt-realtime-2 · Issue #5684 · livekit/agents · GitHub is the tracking issue for first-class support.