Gemini 3.1 Flash Live
#46google · v3.1 · flash live · seit 26. März 2026 · 10× · zuletzt 30. Juni 2026
8
Momentum
Gemini 3.1 Flash Live is Google's native audio-to-audio model for real-time voice interaction with ultra-low latency. It accepts audio, video, images, and text as input and produces audio and text output. Released on March 26, 2026 as a preview, it is architecturally based on Gemini 3 Pro and supports 90+ languages for multimodal live conversations. It is accessible via the Gemini Live API in Google AI Studio, the Gemini API (model ID: gemini-3.1-flash-live-preview), and Vertex AI, and powers the Gemini Live and Search Live consumer products.
Historique du momentum
04.04.03.07.
Fonctionnalités
| Key Benchmark (%) | Scale AI Audio MultiChallenge: 36.1% (with Thinking enabled, leading per Google Blog). ComplexFuncBench Audio: 90.8% (per toolworthy.ai, citing Google's figures). |
| Context Window (Tokens) | 128,000 token input (official per Google DeepMind model card); 64,000 token output (audio & text) |
| Platform | Google AI Studio (aistudio.google.com), Gemini API (gemini-3.1-flash-live-preview), Vertex AI; consumer-side: Gemini Live app (Android/iOS), Search Live |
| Release Date | March 26, 2026 (Preview) |
