Google has introduced two new conversational AI models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, aimed at making interactions with artificial intelligence more natural and efficient. These models allow the AI to see, switch between languages, and continue the conversation while performing tasks, reducing the need for long pauses. The Extended Thinking version is designed to improve the AI's ability to handle complex, multi-step requests, though access to both models is currently limited depending on the service being used. The new models are an upgrade from the Gemini 3.1 Flash Live, introduced in March. Gemini 3.8 Live focuses on responsiveness by analyzing camera input almost instantly, automatically switching between 97 languages, and calling tools or APIs without interrupting the conversation. This allows the AI to confirm requests, explain its progress, and keep the dialogue flowing while handling background tasks. The Extended Thinking version builds on this by adding more advanced reasoning capabilities, making it better suited for complex interactions. Google envisions these models being used in various practical scenarios, such as technical support, booking reservations, onboarding new employees, and even creating fully voice-generated documents. According to Google's Artificial Analysis Speech to Speech Index benchmark, the Extended Thinking model scored 82.6, slightly outperforming similar models from OpenAI and xAI. However, the availability of these models depends heavily on the service being used. Gemini 3.8 Live is rolling out to all users through Search Live, the voice search feature in the Google app on Android and iPhone. Extended Thinking is available in Gemini Live, the voice mode of the Gemini app, and some Workspace services like Gmail and Keep, though access to tools like Google Docs requires a paid AI Pro or Ultra subscription. For developers, both models can be tested in Google AI Studio or through the free API quota. Audio processing costs are based on the amount of audio received and generated, with rates of $0.005 per minute received and $0.018 per minute generated. For companies, deployment is still in a private preview, meaning it is not yet widely available for business use.