Google has launched Gemini 2.5 Deep Think, its most powerful artificial intelligence model so far. The model uses multi-agent systems, which allow several AI agents to work together in parallel. This results in deeper reasoning and more accurate answers, especially for complex tasks.
First unveiled at Google I/O in May 2025, Gemini 2.5 Deep Think is now available to Ultra plan subscribers through the Gemini app. The subscription costs $250 per month. The AI recently helped Google win a gold medal at the International Math Olympiad by solving highly advanced math problems, highlighting its strong problem-solving skills.
Unlike regular AI models, Deep Think takes longer to generate answers, sometimes spending hours reasoning through possibilities. Google has started sharing this version with academic researchers who need detailed and accurate insights. According to Tech Crunch, this long-form reasoning is especially useful in research, coding, and academic development.
In benchmark tests, Gemini 2.5 Deep Think scored 34.8% on Humanity’s Last Exam (HLE), beating competitors like OpenAI’s o3 and xAI’s Grok 4. It also integrates advanced tools like code execution and Google Search to provide rich, creative, and highly detailed responses.
Although running multiple AI agents requires more computing power, Google is funding the infrastructure through premium subscriptions. The company also plans to roll out Gemini 2.5 Deep Think via its Gemini API to select developers and businesses for testing, further strengthening its position in the global AI race.