Google has announced the launch of the “Deep Think” service, which is a new feature within the Gemini 2.5 Pro artificial model, which represents a qualitative shift in the profound thinking capabilities of linguistic models. This new mode allows the model to generate multiple solutions, its analysis, and then combine it to extract the best result, relying on the technique of “parallel multi -agent”.
According to a report published by The Verge, the Deep Think service has proven effective in areas such as solving complex mathematical issues, programming, and strategic thinking, confirming its practical ability and not only theory.
The model showed a remarkable performance in a number of prestigious tests, most notably the International Olympics for Mathematics (IMO) for the year 2025, where he achieved 35 out of 42 points by solving five out of six issues, to obtain an official gold medal from the Coordination Committee. It also outperformed the LiveCodebeench V6 program, bypassing models such as Openai O3 and Grok 4 from Xai.
In the “Humanity’s Last Exam” test (Hle), which measures deep and contemporary thinking, Deep Think achieved a resolution of 34.8%, outperforming its closest competitors.
Technically, the Deep Think depends on the “EXNDEDEDEDEDEDEDEDEDENCE TIME” principle, which gives the model an additional time for internal treatment and analysis before providing the answer. This is associated with the application of augmented learning techniques, which help the model to think of multiple steps and improve its self -performance over time.
The new feature is currently available for Google Ai Ultra subscribers for $ 250 per month, via the Gemini application or through API programming interfaces, and it has begun to launch on a limited scale for research and academic purposes.
According to Deep Think, Deep Think is an advanced step on the path of access to public artificial intelligence (AGI), especially with his ability to contextual, creative and self -acclimatization.
Google confirms that this feature will be pivotal in complex tasks such as designing models, developing algorithms, and scientific research, which makes it a new turning point in the world of artificial intelligence and its practical applications.
