Gemini 3.8 Live brings real-time multimodal voice agents

Google DeepMind's Gemini 3.8 Live and Extended Thinking release supports audio-to-audio dialogue, visual input, tools, 97 languages, API access, asynchronous workflows, spoken reasoning, and SynthID watermarking. New ranking and blind-test coverage adds pricing and human-preference context, while the cited 35.1% banking-task completion rate keeps production reliability unresolved.