Gemini 3.8 Live: What Changes When You Talk to AI in Real Time?
Talking to an assistant is useful only if it can follow interruptions, understand context and keep a task moving. Google introduced Gemini 3.8 Live and 3.8 Live Extended Thinking in September 2026 to improve those parts of a voice conversation. The two variants target different levels of complexity.
What is new in Gemini 3.8 Live?
Google describes Live as a fast, fluid voice model with near-real-time visual grounding. It can take visual input while a conversation continues, switch between supported languages and call tools in the background. Google says it detects transitions among 97 supported languages, a practical improvement for multilingual conversations.
What is Extended Thinking for?
Extended Thinking is designed for more complex, multi-step requests. The model can acknowledge a task verbally, reason while the dialogue continues and narrate progress as it works with tools. It is better matched to a task that needs several steps than to a quick factual question.
How might a person use it?
Google points to troubleshooting in Search Live, longer tasks in the Gemini app and conversational work in Gmail, Docs and Keep. A useful test is to show a device or document and ask a specific question, then add a follow-up without restarting the conversation. For planning, ask the assistant to break the work into steps and explain where it is in the process.
Who can access it?
Google said 3.8 Live was rolling out to developers through the Gemini API and AI Studio, enterprises in private preview, and the public through Search Live. Extended Thinking was rolling out in Gemini Live and to specified Google AI subscribers in Workspace: Pro and Ultra in Docs, and Google AI subscribers in Gmail and Keep. These are separate access routes, so the model name alone does not tell you what your account can use.
Why does real-time context matter?
A conversation feels less useful when each interruption loses the thread or an action freezes the dialogue. Google’s changes aim to let speaking, visual context and background work coexist. The meaningful measure for a reader is whether the assistant can finish an actual task while remaining easy to correct.
In short: Gemini 3.8 Live focuses on fluid conversation; Extended Thinking adds deeper task work. Check the specific product and plan before assuming a feature is in every Gemini surface.
