Google launches Gemini 3.8 Live and 3.8 Live Extended Thinking with real-time reasoning

Google has officially introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, marking the latest evolution in the tech giant's conversational AI capabilities. The newly announced models prioritize natural voice interactions by enabling fluid, uninterrupted dialogue while simultaneously handling complex reasoning tasks. According to the company, these upgrades allow users to speak with Gemini across the Gemini app, Google Workspace, and Search without disrupting ongoing conversations, a significant departure from traditional AI assistants that require pauses during processing.

The two-model architecture serves distinct use cases. Gemini 3.8 Live targets speed and cost efficiency, combining conversational intelligence with real-time visual grounding. Meanwhile, the Extended Thinking variant is engineered for high-complexity workflows, employing multistep reasoning capabilities while maintaining conversational flow. This version introduces early verbal cues like “Let me check that…” to acknowledge user prompts naturally, alongside live progress narration that walks users through background tasks as they unfold. Developers and enterprises can access these models via the Gemini API and Google AI Studio, positioning them as building blocks for production-ready voice agents.

Both models are available starting today, with demonstrations showing applications ranging from real-time chess gameplay using visual context to transforming hand-drawn sketches into functional React components.