Google Unveils Gemini 3.8 Live, Surpassing OpenAI in Voice AI Reasoning

Google has unveiled the Gemini 3.8 Live dual-model system for voice AI, marking an era of real-time reasoning during conversation.

The new models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, are designed to enhance the performance of voice-enabled AI agents.

Gemini 3.8 Live Extended Thinking scored 82.6% on the Artificial Intelligence Speech-to-Speech Index benchmark, narrowly edging out OpenAI's GPT-Live-1 Astra and SpaceXAI's Grok Voice Think Fast 2.0.

Tom Ouyang, a Google engineer, stated that Gemini 3.8 Live Extended Thinking enables users to engage in smoother conversations with Gemini across the Gemini app, Google Workspace, and Search, helping them accomplish complex tasks entirely through voice.