Google has released Gemini 3.1 Flash Live, its latest real-time audio and voice model. It achieves 90.8% on ComplexFuncBench Audio for multi-step function calling and 36.1% on Scale AI's Audio MultiChallenge with thinking enabled. Key improvements include better tonal understanding, acoustic nuance recognition (pitch and pace), and dynamic response adjustment to user emotions. The model powers Gemini Live with faster responses and doubled conversation context length, and enables Search Live's global expansion to 200+ countries in multiple languages. Available to developers via the Gemini Live API in Google AI Studio, to enterprises via Gemini Enterprise for Customer Experience, and to consumers through Gemini Live and Search Live. All audio output is watermarked with SynthID for AI content detection.
Sort: