Thanks — you’re asking which multimodal model gives the lowest latency when used with LiveKit. Good question. The OpenAI Realtime API model is the lowest-latency multimodal option documented for LiveKit integrations. It’s designed for real-time speech and text interactions with end-to-end latencies around ~300 ms, and is optimized for conversational timing. LiveKit’s WebRTC-optimized global transport further reduces transport latency between clients and the model, which helps achieve sub-second responses in real-time voice/text scenarios. If you’re evaluating this for your use case, helpful follow-ups I can assist with: • Do you need guidance on integrating the Realtime API with LiveKit (SDK choices, signaling, or token flow)? • Are you measuring one-way or round-trip latency, and from which geographic regions will your users connect? • Do you want suggestions for tuning (audio codecs, packet sizes, or regional routing) to further minimize latency? Tell me which of those you want next and I’ll walk you through concrete steps. Sources: OpenAI and LiveKit partner to turn Advanced Voice into an API