Google has introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking — two voice models for talking with AI. The company is betting on more natural conversation, work with visual context and tasks that run in the background while the conversation continues.
What the new voice models can do
According to Google, Gemini 3.8 Live can hold conversation context, process visual input in near real time and switch between languages. The model can also run tools and API calls in the background without breaking off the conversation. That matters for voice assistants, where the user doesn’t want to wait in silence after every request.
What Extended Thinking is for
The Extended Thinking version is aimed at harder cases: multi-step tasks, working with services and explaining its progress while it works. Google describes it as an option for voice agents and workflows where a short answer or a single command isn’t enough.
Where Gemini 3.8 Live will appear
The standard version is rolling out in the Gemini API, Google AI Studio and Search Live. Extended Thinking is announced for the API, some enterprise products and Google AI subscriptions, including individual features in Docs, Gmail and Keep. The actual feature set will depend on the country, the plan and the stage of the rollout.
Sources and methodology
Analysis of public sourcesChecked 20 September 2026
- Google blog blog.google
Leave a Reply