Gemini 3.1 Flash Live

Google published the Gemini 3.1 Flash Live announcement on March 26, 2026, credited to product manager Alisa Fortin and developer-relations engineer Thor Schaeff, and labeled the model available in preview via the Gemini Live API. (blog.google) Google reported a 90.8% score for Gemini 3.1 Flash Live on ComplexFuncBench Audio and a 36.1% score on Scale AI’s Audio MultiChallenge with “thinking” enabled, benchmarks used to measure multi-step function-calling and long-horizon reasoning in audio conditions. (blog.google) All audio generated by the model is watermarked using SynthID to allow downstream detection of AI-produced audio. (analyticsindiamag.com) The DeepMind model card states Gemini 3.1 Flash Live is based on Gemini 3 Pro, accepts audio, images, video and text with a token context window up to 128K, and produces audio and text outputs with up to 64K tokens of output. (deepmind.google) Google lists distribution channels for Flash Live as the Gemini App, Google AI Studio, the Gemini API and Vertex/NotebookLM, and says the model will power Gemini Live, Search Live (now rolled out to 200+ countries), and Gemini Enterprise customer‑experience features. (deepmind.google) Pilot integrations and demos named in Google’s announcement include Stitch (voice-driven design workflows) and Ato (an AI companion device for older adults), and enterprise testers cited by Google include Verizon, LiveKit and The Home Depot. (blog.google) Google’s developer pricing pages list Gemini 3.1 Flash‑Lite at $0.25 per 1M input tokens and $1.50 per 1M output tokens, and the Gemini Developer API pricing page enumerates “gemini-3.1-flash-live-preview” as a preview model developers can try in AI Studio. (blog.google)