Gemini 3.8 Live: Improving real-time voice models for voice agents

Google recently introduced the Gemini 3.8 Live version, which optimizes real-time dialogues through parallel reasoning. This improvement aims to increase the efficiency and responsiveness of voice models in real time. The Extended Thinking variant allows systems to simultaneously track multiple thought streams while communicating, increasing the quality and depth of interactions. This development is particularly relevant for voice agents who are able to have complex conversations and interact naturally with users. The integration of such technologies into artificial intelligence systems opens up new opportunities for automating services and improving the user experience. At the same time, the industry is discussing how AI developments should be controlled in the future. Elon Musk, for example, calls for AI laboratories to review each other’s models’ risks to ensure greater security and transparency in dealing with AI. EU Commission President von der Leyen sees dangers from AI differently than US President Trump and emphasizes the need for a more conscious pace of development. These debates underscore the importance of responsible use of AI technologies and their impact on the future of digital communications. In practice, such advances can also be used in integrated business platforms such as xynap to increase the efficiency and performance of voice assistants and voice agents. The AI and voice module integration in xynap enables users to benefit from these technological advancements and optimize their communication processes.