Google Releases Gemini 3.8 Live and Extended Thinking

Google has announced two new model variants: Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking. The pair targets different ends of the AI workload spectrum. The Live models are built for real-time interaction and streaming applications, where low latency is essential. They are positioned for conversational agents, voice assistants, live customer support, and other use cases where users expect quick, natural responses. The Extended Thinking variant takes a different approach. It allocates more compute at inference time, effectively letting the model work through harder problems before producing an answer. That extra reasoning budget can improve performance on multi-step tasks such as complex math, coding, scientific analysis, and strategic planning. The trade-off is speed and cost: deeper reasoning generally takes longer and consumes more resources, so it makes sense for queries where accuracy matters more than instant replies. The release continues Google's rapid cadence of Gemini updates and reflects a broader industry pattern. Major AI labs are increasingly shipping paired models: a fast, low-latency option for everyday interaction and a slower, more deliberate option for difficult problems. This mirrors how developers already choose between lightweight and heavyweight models depending on the task. For businesses, the new variants could make it easier to build products that blend speed and depth. A customer service bot might use Gemini Live for immediate responses, then hand off complex cases to Extended Thinking. A coding assistant could offer quick suggestions while using deeper reasoning to debug tricky issues. The main questions will be pricing, availability, context limits, and how well the models perform in real-world tests. If they deliver, Google will strengthen its position in an increasingly crowded market for real-time and reasoning-focused AI systems.