Google brings background tool use and deeper reasoning to live voice AI
Google introduced Gemini 3.8 Live and Live Extended Thinking. The models combine conversation with visual context and background tool calls; Extended Thinking targets more complex, multi-step work. Developer access is rolling out through the Gemini API and AI Studio, while enterprise availability includes private previews. The announcement's demonstrations and performance claims are Google's, not independent evaluations.
Context & source check
Voice becomes more useful when it can complete a bounded task, not just answer a question. For a workflow pilot, test interruptions, incorrect tool calls and confirmation before consequential actions alongside conversational quality.
Model rollout. Company source. Source checked Sep 16, 2026. Not independently tested.