Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking
Google is putting its new Gemini Live models into voice workflows that can keep talking while they reason, see context, and run tools in the background.
Gemini 3.8 Live is framed as the faster, lower-cost model for fluid voice agents with visual grounding and support for switching among 97 languages mid-conversation. Gemini 3.8 Live Extended Thinking is the higher-reasoning version, meant for multi-step tasks where the model can narrate progress while background actions continue. Google says both are rolling out through the Gemini API and Google AI Studio, with enterprise previews and consumer availability split across Search Live, Gemini Live, Workspace, Gmail, Keep, and Docs. The company also says AI-generated audio from these products carries SynthID watermarking. Google DeepMind's note
Gemini 3.8 Live is framed as the faster, lower-cost model for fluid voice agents with visual grounding and support for switching among 97 languages mid-conversation. Gemini 3.8 Live Extended Thinking is the higher-reasoning version, meant for multi-step tasks where the model can narrate progress while background actions continue. Google says both are rolling out through the Gemini API and Google AI Studio, with enterprise previews and consumer availability split across Search Live, Gemini Live, Workspace, Gmail, Keep, and Docs. The company also says AI-generated audio from these products carries SynthID watermarking. Google DeepMind's note
score 8