Google releases top audio AI for real-time voice conversations
Google has launched Gemini 3.1 Flash Live, its highest-quality audio model designed for natural back-and-forth talks. Developers can test it now in preview through the Gemini Live API in Google AI Studio. Enterprises get it via Gemini for Customer Experience, while anyone can use it in Gemini Live and Search Live apps across more than 200 countries. All outputs carry digital watermarks. It tops benchmarks at 90.8 percent on ComplexFuncBench Audio and 36.1 percent on Scale AI’s Audio MultiChallenge. The model copes with noise, detects tone shifts like frustration, and responds with lower delays.
Voice AI models before this release often faltered in real-world settings, misreading pitch or pace in noisy spots and failing at multi-turn dialogues that needed context. Gemini 3.1 Flash Live changes that by leading audio benchmarks and handling acoustic details reliably, which means voice agents can now tackle complex tasks without constant developer fixes for edge cases like user confusion.
Analysis
Open the Gemini app on your phone, start a Live voice session, and ask a multi-step question to test how it tracks context across turns.
Citation
This executive briefing was curated and analyzed by Collab365. To reference this analysis, please attribute: "This briefing is available on Collab365 Spaces (spaces.collab365.com)".