Google introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, two models designed to bring near real-time reasoning to voice agents and spoken AI interactions. Both are rolling out starting today across selected Google products and developer services, with enterprise access initially limited to private previews in several offerings.
Gemini 3.8 Live is positioned as the scale- and cost-efficient model, combining conversational intelligence, fluid dialogue and visual grounding. Extended Thinking is intended for higher-complexity work requiring increased intelligence and multi-step reasoning. Google says the models provide building blocks for production-ready voice agents while supporting more fluid voice interactions in the Gemini app, Google Workspace and Search.
According to Google, Gemini 3.8 Live Extended Thinking ranked first overall on Artificial Analysis’ Speech to Speech Quality Index with a score of 82.6. The company also reported scores of 68.6% on τ-Voice, 35.1% on Sierra’s τ-Voice-banking benchmark and 97.7% on Big Bench Audio. Gemini 3.8 Live placed second in the Speech Agent Arena, while Google said both models advanced the Pareto Frontier on ServiceNow’s EVA-Bench by balancing accuracy and conversational quality in complex workflows.
Gemini 3.8 Live can process visual inputs in near real time and automatically detect and switch among 97 supported languages during a conversation. It can also execute tools and API calls in the background while continuing to speak with the user. For deeper tasks, Extended Thinking is designed to reason and speak simultaneously, using acknowledgments such as “Let me check that…” and spoken progress updates while multi-step work continues.
The Gemini Live API is used by developer platforms including Agora, Fishjam, LiveKit, Pipecat, Vercel and Vision Agents, which manage real-time media-streaming infrastructure for voice interfaces. Google also identified Salesforce, Genspark and Lumeris as partners interested in the models’ latency, conversational fluidity and tool-calling capabilities. The company said all audio generated by its AI products is watermarked with SynthID, an imperceptible marker intended to keep AI-generated content detectable and help prevent misinformation.
Availability
Gemini 3.8 Live is rolling out through the Gemini API and Google AI Studio, in private preview for Gemini Enterprise, and in Search Live. Google said it is coming soon to Gemini Enterprise for Customer Experience. Extended Thinking is also rolling out through the API and AI Studio, in private preview for Gemini Enterprise, and through Gemini Live. It is available to Google AI Pro and Ultra subscribers in Workspace in Docs and to all Google AI subscribers in Gmail and Keep, with planned availability for Gemini Enterprise for Customer Experience and Google Workspace business customers.