Google DeepMind has launched Gemini 2.5 Deep Think, its most advanced AI reasoning model, now available exclusively to Google AI Ultra subscribers through the Gemini app. This iteration is a variation of the model that earned a gold medal at the 2025 International Mathematical Olympiad (IMO), demonstrating exceptional capabilities in complex problem-solving, including mathematics, coding, and scientific reasoning. The model employs parallel thinking and reinforcement learning techniques to brainstorm multiple solutions simultaneously, enabling it to tackle challenging tasks such as mathematical conjectures, advanced coding problems, and strategic planning. Performance benchmarks for Gemini 2.5 Deep Think include a 34.8% score on Humanity’s Last Exam (HLE) without external tools, 86.6% on Live Code Bench, and 99.2% on the American Invitational Mathematics Examination (AIME) 2025. The AI is accessible to subscribers paying $250 per month and is currently being used by mathematicians and researchers to explore its potential. Google CEO Sundar Pichai confirmed the rollout and highlighted the model’s ability to assist with complex scientific literature and iterative development. The release is part of an intensifying competition in the AI field, with OpenAI expected to launch GPT-5 soon, which is anticipated to offer improvements in coding, math, and agent control. Google’s Gemini 2.5 Deep Think represents a notable advancement in AI reasoning technology, emphasizing creative problem-solving and multi-threaded inference.