• Is there a recommended low-latency Gemini model or config specifically for real-time phone/SIP conversations? • Are there best practices to reduce Gemini Realtime latency (buffer sizes, proactivity, thinking config, audio chunking)? • Google mentions Gemini 3 Flash as optimized for ultra-low latency — is there any plan or workaround to use it with LiveKit Agents (v1.3.6)?

Thanks for reaching out, harsha. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

yes

Thanks — glad that matches. Does the information meet your needs, or would you like more detail on any of these? You can ask for: • Specific config examples for LiveKit Agents (JSON/YAML or code snippets) • Recommended WebRTC frame sizes, chunking, and VAD settings for phone/SIP flows • How to enable proactivity and

thinking_config

for a native audio model • Options for deployment/network placement to minimize transit latency

how to enable gemini 3 flash intour model is it possible

Regarding enabling Gemini 3 Flash in LiveKit Agents (v1.3.6): • Gemini 3 Flash is not currently supported or listed for use with LiveKit Agents v1.3.6. You cannot enable it in that release. • What you can use today: ◦

gemini-2.5-flash

via

RealtimeModel

for low-latency voice use. ◦ The experimental

gemini-2.5-flash-native-audio-preview-09-2025

for native audio + proactivity features (recommended for real-time phone/SIP flows). • Next steps I can help with: ◦ Provide a sample LiveKit Agents config or code snippet for

gemini-2.5-flash

or the native-audio preview. ◦ Show recommended WebRTC frame sizes, chunking, VAD and

thinking_config

settings to minimize latency. ◦ Discuss deployment/network placement to reduce transit time. Which of the above would you like next?