Technology

Google Launches Gemini 3.8 Live and Extended Thinking Models for Voice and Audio Interaction

By Tom Ouyang

Google today announced the release of two new Gemini models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, designed to improve real‑time reasoning for voice agents and conversational AI. Gemini 3.8 Live is optimized for scale and cost efficiency, combining conversational intelligence with fluid dialogue and visual grounding. Gemini 3.8 Live Extended Thinking targets high‑complexity tasks, offering increased intelligence and multi‑step reasoning.

The models are positioned as production‑ready building blocks for developers and enterprises, enabling reliable voice agents and more intuitive interactions across the Gemini app, Google Workspace, and Search. Gemini 3.8 Live Extended Thinking achieved the top overall score on Artificial Analysis’ Speech to Speech Quality Index (82.6) and led agentic task completion with 68.6% on the τ‑Voice benchmark and 35.1% on Sierra’s τ‑Voice‑banking benchmark. It also scored 97.7% on the Big Bench Audio reasoning test while maintaining a competitive price.

Gemini 3.8 Live secured second place in the Speech Agent Arena and remains cost‑effective for large‑scale deployment. On ServiceNow’s EVA‑Bench, both models demonstrate a balance of accuracy and conversational quality, pushing the Pareto frontier for complex workflows.

Key capabilities include near‑real‑time visual input processing, automatic language detection across 97 supported languages, and background tool and API execution that allows the model to acknowledge requests while tasks complete. Gemini 3.8 Live Extended Thinking can reason and speak simultaneously, using verbal cues such as “Let me check that…” to maintain conversational flow while narrating progress on multi‑step tasks.

Developer platforms such as Agora, Fishjam, LangChain, LiveKit, Pipecat, Vercel, and Vision Agents can use the Gemini Live API to build high‑performance voice‑driven interfaces, while partnerships with Salesforce, Genspark, and Lumeris highlight the models’ low latency, fluidity, and tool‑calling capabilities.

All audio generated by the models is watermarked with SynthID, an imperceptible watermark embedded directly into the audio to ensure AI‑generated content remains detectable.

Availability: Gemini 3.8 Live is rolling out today for developers via the Gemini API and Google AI Studio, for enterprises in private preview in Gemini Enterprise, and for all users in Search Live. Gemini 3.8 Live Extended Thinking is also rolling out today for developers via the Gemini API and Google AI Studio, for enterprises in private preview in Gemini Enterprise and soon for Google Workspace business customers, and for all users in Gemini Live, Google AI Pro and Ultra subscribers in Workspace Docs, and all Google AI subscribers in Gmail and Keep.