Google Preps Gemini 3.1 Preview
Google is preparing a Gemini-3.1-Flash-Live-Preview for Vertex AI and AI Studio, signaling possible upgrades to Gemini Live and tighter integration into cloud ML tooling. The preview suggests faster model iteration paths for developers building on Google Cloud. (x.com)
Google published the gemini-3.1-flash-live-preview entry in its Gemini API changelog on March 26, 2026. (ai.google.dev) The model is described as an audio-to-audio (A2A) low-latency model optimized for real-time dialogue with acoustic-nuance detection, numeric precision, and multimodal awareness. (ai.google.dev) Developers can invoke Gemini 3.1 Flash Live through the Gemini Live API inside Google AI Studio for interactive testing and prototyping. (blog.google) Enterprises can integrate the same Live-capable surface via Vertex AI’s Live API to embed low-latency voice and vision agents into production services. (docs.cloud.google.com) Google has rolled Gemini 3.1 Flash Live into Search Live and Gemini Live to enable voice-plus-camera conversations in more than 200 countries and territories where AI Mode is available. (blog.google) DeepMind’s model card lists official distribution channels — Gemini App, Google AI Studio, Gemini API, Google Antigravity, and NotebookLM — and documents that the model can generate both audio and text outputs. (deepmind.google) The Gemini Live API documentation states the Live API processes continuous streams of audio, images, and text to deliver immediate, human‑like spoken responses for low‑latency conversational agents. (ai.google.dev) Google and DeepMind note built-in safety features for Flash Live, including SynthID audio watermarking for generated speech and multilingual support that underpins the global Search Live expansion. (deepmind.google)
