Google recently launched a “Deep Thinking” mode in its Gemini 3 Pro AI model. This mode, currently only available to Google AI Ultra subscribers, allows the model to process complex issues in mathematics, science, and logic by exploring multiple hypotheses simultaneously, as Google mentioned in its blog.
The “Deep Thinking” mode also helps in producing more accurate code and improving visualization and prototyping processes. It also offers superior performance compared to the older Gemini 2.5 model, achieving high scores in key evaluation tests:
– Scored 41% on Humanity’s Last Exam
– and 93.8% on GPQA Diamond
– and 45.1% on ARC-AGI-2, which specializes in code execution
After a brief delay last month for additional safety testing, the mode is now available to Google AI Ultra subscribers, which costs $250 per month.
To use this feature, the Deep Thinking mode must be selected from the tools menu within Gemini 3, keeping in mind that answers may take a few minutes to appear.
