Google rolls out Gemini voice models that speak while reasoning

A voice assistant need not fall silent while it gets something done. Google says Gemini 3.8 Live can keep talking while running tools and API calls in the background, process visual inputs and switch between 97 supported languages.

In their announcement, Tom Ouyang and Malini Jaganathan describe two models: a standard version aimed at scale and cost efficiency, and Extended Thinking for more complex, multi-step work. Google says the latter reasons and speaks simultaneously, providing progress updates as tasks proceed.

Where the rollout starts

Both models are rolling out through the Gemini API and Google AI Studio. Standard Live is also reaching Search Live; Extended Thinking is reaching Gemini Live.

Workspace access varies by subscription: Extended Thinking is rolling out in Docs for Google AI Pro and Ultra subscribers, and in Gmail and Keep for all Google AI subscribers. Both models enter private preview in Gemini Enterprise, with customer-experience access planned.

Google reports a 68.6% task-completion score for Extended Thinking on τ-Voice, but 35.1% on its banking counterpart. Those benchmark results support testing it on particular workflows, not assuming that fluent conversation means dependable completion.