Gemini 3.8 Live and 3.8 Live Extended Thinking: A practical guide to Google’s newest live dialogue models

Google introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking as voice-first dialogue models meant to make spoken interaction with software more natural and useful. Both models process audio in near real time, accept visual inputs to add context to responses, and can execute tools or API calls in the background while continuing the conversation.

Gemini 3.8 Live is positioned for scale and cost efficiency. It aims to deliver fluid conversational quality, quick turn-taking, and real-time visual grounding appropriate for applications that need many simultaneous users or lower latency.

Gemini 3.8 Live Extended Thinking emphasizes higher intelligence and deeper multi-step reasoning. It's intended for workflows that need more complex, agentic behavior: the model can begin speaking to acknowledge or narrate progress while it continues multi-step tasks in the background.

• Seamless interruptions and language switching: The models can handle interruptions naturally and automatically switch between 97 languages mid-conversation. This reduces friction for multilingual or noisy environments.
• Visual grounding: Both models can process live visual inputs to inform responses, enabling scenarios like live troubleshooting, guided onboarding, or interpreting a user's camera feed.
• Background execution: The models execute tool calls and API requests asynchronously. They can acknowledge a request verbally and keep the user engaged while work proceeds behind the scenes.
• Live narration and early cues: Extended Thinking gives verbal cues like "Let me check that…" and narrates progress on long-running tasks, so users don't lose conversational continuity when the system is executing complex operations.

Developer and enterprise integration

For product teams building conversational agents, the updates reduce the need to suspend dialogue while executing actions. You can design voice flows that acknowledge user intent quickly, keep users informed with live progress narration, and incorporate visual context for richer responses. For enterprise use, Extended Thinking is presented as the option for workflows that require stronger reasoning and task orchestration.

Google lists availability via the Gemini API, the Gemini app, and Workspace integrations. The announcement included demo scenarios and referenced developer resources to build real-time voice applications with these models.