Gemini 3.8 Live Extended Thinking: First Place in Voice AI Quality

Google has released two voice AI models: Gemini 3.8 Live and 3.8 Live Extended Thinking. The Extended Thinking variant took first place on the Artificial Analysis Speech to Speech Quality Index with a score of 82.6, scored 97.7% on Big Bench Audio, and supports 97 languages with automatic switching mid-conversation.

As a vibe coder, I see this as a ready-made stack for voice agents — LiveKit and Pipecat are already integrated, so there's no need to stitch together five different APIs. Extended Thinking reasons and speaks at the same time, which makes the agent feel far less like a black box. It's a direct competitor to GPT-Live-1, and for now Google holds the top spot on quality.