Google launches Gemini 3.8 Live voice AI models for real-time conversation

Google has launched two new voice AI models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, designed to enable near-real-time conversation and background task execution. The company described them as its most advanced voice processing models to date, capable of maintaining a natural conversational pace while simultaneously performing complex operations. The models are engineered to execute tool and API calls in the background, allowing AI agents to continue interacting with users while working on assigned tasks. This approach aims to create a more seamless and human-like experience, avoiding the constant interruptions that can occur when an AI system pauses to perform an action. Gemini 3.8 Live is positioned for high-volume, cost-effective voice interfaces, while the Extended Thinking variant is built for more demanding reasoning and can orchestrate multiple agents to solve background tasks. Both models support automatic language detection and can switch between 97 languages mid-conversation. They also feature near-real-time visual grounding and use early verbal cues like "let me check that" to acknowledge prompts naturally.