Deep Think is available in the Google Gemini App. How to try it.

Published: Friday, August 1, 2025 at 2:07 pm

Google Rolls Out Deep Think AI Model for Subscribers

Google is expanding access to its advanced AI reasoning model, Deep Think, within the Gemini app. The model, designed for complex problem-solving, will be available to subscribers of Google AI Ultra, a premium subscription service costing $250 per month. Access to Deep Think begins on Friday, August 1.

Deep Think was initially unveiled at Google I/O 2025. The model is touted as Google's most sophisticated AI to date, boasting enhanced reasoning capabilities and the ability to tackle intricate problems using parallel thinking techniques. Google claims Deep Think delivers "more detailed, creative, and thoughtful responses" to user prompts.

The company is positioning Deep Think as a leader in the AI landscape, citing its performance on benchmark tests. Google's model achieved a score of 34.8% on Humanity's Last Exam (HLE), a complex AI assessment encompassing math, science, and humanities. This score surpasses the performance of competing models, including xAI's Grok 4, which scored 25.4%, and OpenAI's o3, which scored 20.3%. Deep Think also demonstrated superior performance on LiveCodeBench6, a test focused on coding tasks.

Furthermore, Google is releasing an official version of Deep Think, which achieved a gold-medal standard at the International Mathematical Olympiad (IMO) in July. This specialized version will be provided to a select group of "trusted testers," including mathematicians and academics. The president of the IMO praised Deep Think's solutions as "astonishing in many respects."

BNN's Perspective:

Google's continued investment in AI, particularly in advanced reasoning models like Deep Think, is a positive development. While the high cost of the subscription may limit access for some, the potential for innovation and advancements in complex problem-solving is significant. The company's commitment to rigorous testing and validation, as evidenced by its performance on benchmark tests and the IMO, is encouraging. However, it is important to monitor the long-term impact of these advanced AI models and ensure responsible development and deployment.

Keywords: Google, Gemini, Deep Think, AI, Artificial Intelligence, Reasoning Model, Google AI Ultra, Subscription, Complex Problem Solving, HLE, Humanity's Last Exam, IMO, International Mathematical Olympiad, Coding, xAI, Grok 4, OpenAI, o3, Gemini app

Full Story