Google introduced the Gemini 3.8 Live and Extended Thinking models

Company Google released two new versions of neural networks: Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking. The models are focused on working with voice interfaces and complex tasks that require in-depth data analysis in real-time.

The key feature of the updated line is the support of visual grounding, which allows models to relate text responses to visual data. The architecture integrates SynthID technology for automatically applying digital watermarks to generated content, ensuring the identification of AI materials.

The Extended Thinking version received improved reasoning mechanisms, which allow the system to process multi-step logical chains before providing a final answer. The tools are already available to developers through the Gemini API and Google AI Studio development environment.

The implementation of these models reduces delays in processing voice requests, bringing interaction with AI closer to natural human communication. The update is aimed at improving the accuracy of responses in scenarios where not only fast reaction but also in-depth context processing is required.