Google is launching Gemini 2.5 Deep Think within the Gemini app for Google AI Ultra subscribers. As of Friday, August 1, those enrolled in the $250 monthly AI plan will gain access to Deep Think. Initially revealed at Google I/O 2025, this model is tailored for advanced math and programming, delivering enhanced reasoning skills and parallel thinking methods. Google asserts that Deep Think offers more intricate and imaginative replies to inquiries.

Gemini 2.5 Deep Think is reported to excel beyond other AI models, utilizing scores from Humanity’s Last Exam (HLE) as a reference point. It achieved a score of 34.8% on the assessment, whereas xAI’s Grok 4 scored 25.4% and OpenAI’s o3 reached 20.3%. It also outperformed both xAI and Anthropic models on LiveCodeBench6, a coding challenge evaluation.

Google has confirmed the rollout of the official edition of Gemini 2.5 Deep Think, which earned a gold-medal distinction at the International Mathematical Olympiad (IMO) in July. This iteration will be accessible to an exclusive group of “trusted testers,” comprising mathematicians and scholars. The IMO president praised Deep Think’s solutions as “remarkable in numerous ways.”