PANews reported on December 5th that Gemini has officially launched the "Gemini 3 Deep Think" mode, its most powerful multi-hypothesis parallel inference feature to date. This mode generates more detailed and complex code output through iterative inference loops, suitable for visualization, prototyping, and experimental needs. Currently, this feature is available to Google AI Ultra subscribers; users can select the "Thinking" model and enable "Deep Think" mode in the Gemini app.