Gemini has launched its "Gemini 3 Deep Think" mode, enhancing its capabilities in reasoning and code generation. This new feature, described as the most powerful multi-hypothesis parallel inference tool, allows for the creation of detailed and complex code outputs through iterative inference loops. It is designed to support visualization, prototyping, and experimental applications. Currently, the "Deep Think" mode is accessible to Google AI Ultra subscribers, who can activate it by selecting the "Thinking" model within the Gemini app.