Google is rolling out its latest and most advanced AI model, Gemini 2.5 Deep Think, today, although access will be limited to a select group of users. This powerful AI was first introduced during the I/O conference in May and is now available exclusively through the Gemini app. Designed to tackle the most intricate queries, Deep Think requires significantly more computational resources than its predecessors, making it accessible only to subscribers of Google’s premium 0 AI Ultra plan.
Enhanced Analytical Capabilities
Deep Think builds upon the foundation of Gemini 2.5 Pro, enhancing its capabilities by increasing “thinking time” through greater parallel analysis. This innovative approach allows the AI to explore various problem-solving methods, revisiting and remixing hypotheses to produce higher-quality outputs.
Similar to other sophisticated tools within the Gemini suite, Deep Think requires several minutes to formulate a response. This extended processing time contributes to its proficiency in areas such as design aesthetics, scientific reasoning, and coding. Google has subjected Deep Think to a comprehensive series of benchmarks, demonstrating its superiority over both the standard Gemini 2.5 Pro and competing models, including OpenAI’s o3 and Grok 4.
One notable achievement of Deep Think is its performance on Humanity’s Last Exam, a challenging set of 2,500 complex, multi-modal questions spanning over 100 subjects. While other models typically score between 20 to 25 percent, Gemini 2.5 Deep Think has impressively achieved a score of 34.8 percent, showcasing its advanced analytical prowess.
