Gemini 3.1 Flash Live

2026-03-26 · Publisher : Google · Domain : Artificial Intelligence

Google launches Gemini 3.1 Flash Live, its most accomplished audio and voice model, for real-time voice agents, through the Live API, Search Live and Gemini Live, in more than 200 countries and territories.

What changes

Better multi-step function following in audio: 90.8% on ComplexFuncBench Audio.
Better understanding of tone (pitch, rhythm, user frustration or confusion).
Gemini Live follows the thread of a conversation twice as long.
Natively multilingual model, which enables the extension of Search Live to more than 200 countries and territories.
Audio output marked with SynthID.

### Specifications - Availability: Live API in Google AI Studio (preview), Gemini Enterprise for Customer Experience, Search Live and Gemini Live. - Identifier and context window: not disclosed in the post. ### Pricing Grouped line from the current pricing page for the Live 3.1 and 3.8 models: input 0.75 USD (text) or 0.005 USD per minute (audio); output 4.50 USD (text) or 0.018 USD per minute (audio). Launch pricing not verified. ### Benchmarks | Benchmark (scores published by the vendor) | 3.1 Flash Live | |---|---| | ComplexFuncBench Audio | 90.8% | | Audio MultiChallenge (with extended thinking) | 36.1% | ### Key takeaways A next-generation voice model for customer service agents and voice assistants, still in preview for developers.

Links

Tags : Audio, Google, Gemini, Live, Voix