Google Launches Gemini 3.8 Live for Real-Time Voice AI
View original source →Google released two new real-time voice models this week—Gemini 3.8 Live and a more deliberate "Extended Thinking" version—built for natural, low-latency spoken conversations with AI agents that can also reason through harder requests using tools mid-conversation.
Key points:
• Gemini 3.8 Live is built for fast, natural back-and-forth speech, while the Extended Thinking variant trades some speed for the ability to reason more carefully and use tools before responding
• The release targets developers building voice assistants, customer-service agents, and other live conversational products rather than being a consumer-facing chat feature
• The launch continues Google's push into agentic, tool-using models across both text and voice interfaces
Real-time voice AI has lagged behind text-based chat in naturalness. A model built specifically to reason and use tools while staying conversationally fast is a meaningful technical step for products that depend on spoken interaction.
Why It Matters: Teams building voice-based customer or employee tools should evaluate Gemini 3.8 Live directly against their current voice AI for latency and task-completion accuracy.