Google Unveils Olympiad-Winning Gemini 2.5 Deep Think AI for Public Use – But With a Limitation

Google has made waves in the AI community by releasing its highly acclaimed Gemini 2.5 Deep Think model to the public, a system that recently achieved a gold medal standard at the International Mathematical Olympiad (IMO). This marks a significant milestone in AI reasoning, showcasing the model's ability to tackle complex mathematical problems with human-like proficiency.

However, there's a notable catch. The version of Gemini 2.5 Deep Think now available to users is not the same high-performing model that clinched the Olympiad victory. Instead, Google has rolled out a faster but less powerful variant, downgraded to what some sources describe as a 'bronze' level for general accessibility.

This public version is accessible through Google's premium subscription plan, Google AI Ultra, targeting users who require advanced reasoning capabilities for tasks in mathematics, science, and coding. Despite the downgrade, the model still offers impressive performance, particularly in areas like debugging and problem-solving.

Google's decision to release a scaled-down version likely stems from computational constraints and the need to balance performance with user accessibility. The original competition model, which could process solutions over extended periods, demanded significant resources that may not be feasible for widespread use.

The launch of Gemini 2.5 Deep Think underscores Google's commitment to pushing the boundaries of AI innovation, even as it navigates the challenges of scaling such advanced technology. The company continues to refine its multi-agent architecture, which allows the AI to test multiple ideas in parallel, enhancing its problem-solving efficiency.

As the AI landscape evolves, this release positions Google as a frontrunner in reasoning-focused AI, though users eager for the full gold-medal experience may need to wait for further updates or expanded access to the top-tier model.