Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking arrived Sep 15, 2026, as Google DeepMind’s most advanced live dialogue models, built to reason, see and call tools without pausing the conversation.
The pitch is simple: talk never stops.
According to Deepmind, the standard 3.8 Live is tuned for scale and cost efficiency with fluid dialogue and visual grounding, while Extended Thinking adds multi-step reasoning and speaks while it works, opening with cues like "Let me check that…" and narrating progress live. The team of Tom Ouyang, Principal Engineer, and Malini Jaganathan, Member of Technical Staff, put numbers on it: Extended Thinking took the #1 spot on Artificial Analysis’ Speech to Speech Quality Index at 82.6, hit 68.6% on τ-Voice and 35.1% on Sierra’s τ-Voice-banking, and scored 97.7% on Big Bench Audio. Gemini 3.8 Live took second place in the Speech Agent Arena and both pushed the Pareto Frontier on ServiceNow’s EVA-Bench for balancing accuracy with conversational quality.
What had to be true before this was already in motion. Google needed a Live API that could stream audio and video in near real time, handle 97 languages with mid-sentence switching, and run tool calls in the background. It also needed distribution: Gemini app, Google Workspace and Search Live where these models would actually live. The Sep 15 post was updated Sep 17, 2026, but the developer positioning was clarified off-blog, where Google lists Gemini 3.8 Live as the default option for most low-latency voice agent experiences and real-time dialogue without reasoning-induced delays.
What still has to happen is rollout. Both models are live now in the Gemini API and Google AI Studio. For enterprises they are in private preview in Gemini Enterprise, with Customer Experience and Google Workspace business access flagged as coming soon. For consumers, 3.8 Live is in Search Live and Extended Thinking is in Gemini Live for Google AI Pro and Ultra subscribers, plus Docs Live, Gmail Live and Keep Live. The demos point to that future: onboarding with live visual context, playing chess from a camera feed, turning sketches into React components on voice feedback, and chaining bookings through asynchronous function calls.