Gemini 3.1 Flash Live: Google's Real-Time Voice Model for AI Agents

Google launched Gemini 3.1 Flash Live on March 26, its highest-quality audio model designed for building real-time voice and vision agents. The model processes the world around it — audio, video, and tool calls — and responds at conversational speed with lower latency than its predecessor, 2.5 Flash Native Audio.

For the agentic ecosystem, the critical feature is native tool use during live audio sessions. Agents can now see, hear, and act simultaneously — querying databases, calling APIs, or controlling software while maintaining a natural voice conversation. The model also better distinguishes relevant speech from background noise (traffic, television) and recognizes acoustic nuances like pitch and pace.

Gemini 3.1 Flash Live is available through the Gemini Live API in Google AI Studio and supports over 90 languages for real-time multimodal conversations. Google is using it to power Search Live globally across 200+ countries.

https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-flash-live/