Google Launches Deep Think Administrative Model For Ultra AI Subscribers

Google launched Deep Think for Google AI Ultra subscribers (photo: dock. Google)

JAKARTA Google announced the launch of Deep Think mode for Google AI Ultra subscribers. This feature was previously only available in preview mode, along with the launch of Gemini 3 Pro last month.

Deep Think is referred to as Google's most advanced reasoning mode that can solve complex problems. This mode utilizes high-level parallel reasoning techniques to explore various hypotheses simultaneously.

With this recurrent reasoning process, the system can produce smoother and more nuanced codes. As a result, this feature can be used for better data visualization and more effective prototype manufacturing in various fields.

Deep Think's ability is claimed to be relevant for dealing with complicated math, science and logic issues. However, because of its high workload, Deep Think takes up to a few minutes to provide answers.

Deep Think is based on Gemini 2.5 which was introduced last August. The Artificial Intelligence (AI) model shows a significant increase in performance in some of the industry's key benchmarks.

At the Humanitarian Final Test benchmark, Deep Think scored 41.0 percent without using external tools. Meanwhile, at GPQA Diamond, this feature recorded an impressive result of 93.8 percent.

Google AI Ultra subscribers with a subscription fee of US$250 (IDR 4.1 million) per month can now access Deep Think. This mode can be selected from the Tool menu when the Thinking option is set in the drop-down menu of the model.