Google has recently unveiled two groundbreaking models: Gemini3.8Live and Gemini3.8Live Extended Thinking. These innovative models boast a suite of core functionalities, including near-real-time reasoning capabilities, the ability to process voice and thought simultaneously, seamless background tool and API integration, automatic detection and switching among 97 languages, and near-instantaneous visual localization. Additionally, they keep users informed about their thought processes through intuitive language prompts. Currently, these models are accessible on relevant developer platforms, catering to a wide range of application scenarios and demonstrating exceptional performance across various tests. Google has forged partnerships with multiple platforms to streamline the integration process for developers, ensuring ease of use. Moreover, all audio generated by these models is embedded with SynthID watermarks, enhancing security and traceability. This significant upgrade propels real-time voice interaction beyond mere 'listening and speaking,' elevating it to a sophisticated level of understanding, reasoning, and task execution. It marks a pivotal step towards the development of voice AI that enables continuous, multi-tasking intelligent agent interaction, setting a new standard in the field.