Google (GOOGL.US) Releases Gemini 3.8 Live Dual Model: Voice AI Enters the “Reasoning and Dialogue” Era, and Measured Scores Surpass OpenAI and SpaceXAI

Zhitongcaijing · 2d ago

The Zhitong Finance App learned that Google (GOOGL.US) released two new Gemini models on Tuesday — Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, which focus on improving voice intelligence performance.

Google engineer Tom Ouyang said in a blog post: “For developers and businesses, these models provide the basic module for building reliable, production-ready voice agents. They also make conversations with Gemini smoother and more collaborative through the Gemini app, Google Workspace, and search—helping users complete complex tasks using just their voice.”

According to information, Gemini 3.8 Live Extended Thinking scored 82.6% on the AI Speech to Speech Evaluation Benchmark (Artificial Intelligence Speech to Speech Index), surpassing OpenAI's GPT-Live-1 Astra (81.5%) and SpaceXAI (SPCX)'s Grok Voice Think Fast 2.0 (81.3) by a narrow margin %).

image.png

Ouyang said, “For tasks that require deep reasoning, 3.8 Live Extended Thinking can talk while reasoning. It brings a greater level of intelligence to complex workflows while keeping conversations flowing — responding naturally to users through early voice prompts such as 'Let me check... ', and guiding users through the progression of multi-step back-office tasks with real-time progress narratives.”

Additionally, Google is collaborating with CRM.US (CRM.US), Genspark, and Lumeris to explore application scenarios for the new model.

Bob Van Osten, vice president of Agentforce products at SAFTSE, said, “By combining real-time multi-modal capabilities with intelligent AI, we are exploring new ways to create a richer and more intuitive customer experience from initial contact to problem solving.”

image.png

According to reports, Gemini 3.8 Live is now open to developers. It can be used through Gemini API and Google AI Studio, private preview can be opened in Gemini Enterprise, and all users can log in to Search Live at the same time. Gemini 3.8 Live Extended Thinking is now open to developers. It can be used via Gemini API and Google AI Studio to open private previews in the enterprise version of Gemini Enterprise, and is available to all users on Gemini Live and Google AI Pro.