This report details the imminent release of Google's groundbreaking AI model, Gemini 2.5 Deep Think, to its premium subscribers. It explores the model's advanced capabilities in tackling complex reasoning tasks, its impressive performance against competitor AIs in rigorous benchmarks, and the exclusive access granted to a select group of experts for its specialized version.
Unlocking Unprecedented AI Insight: Deep Think Arrives for Elite Minds
Google's Advanced AI Model Now Accessible to Premium Users
Google is rolling out its sophisticated AI reasoning model, Gemini 2.5 Deep Think, within the Gemini application. This cutting-edge technology will be available to subscribers of Google AI Ultra, a premium service priced at $250 per month, beginning Friday, August 1st.
Deep Think: A Leap in AI Reasoning and Problem Solving
First introduced at Google I/O 2025, Deep Think is specifically designed to handle complex mathematical and coding problems. It stands as Google's most advanced AI model yet, distinguished by its enhanced reasoning capabilities and its innovative use of parallel thinking techniques to derive solutions. Google's official announcement highlights Deep Think's capacity to provide "more detailed, creative, and thoughtful responses" to a variety of prompts.
Outperforming Rivals: Deep Think's Benchmark Achievements
Google asserts that Gemini 2.5 Deep Think surpasses other leading AI models, citing its performance in Humanity's Last Exam (HLE) as a key indicator. This challenging AI assessment, encompassing questions from mathematics, science, and humanities, saw Google's model achieve a score of 34.8%. In comparison, xAI's Grok 4 scored 25.4%, and OpenAI's o3 reached 20.3%. Furthermore, Deep Think also demonstrated superior performance against AI models from xAI and Anthropic in LiveCodeBench6, a test focused on coding tasks.
Exclusive Access for Elite Minds: The Official Release of Gemini 2.5 Deep Think
Google has also announced the release of the official version of Gemini 2.5 Deep Think, a model that garnered a gold medal standard at the International Mathematical Olympiad (IMO) in July. This highly specialized version will be made available to an exclusive group of "trusted testers," including prominent mathematicians and academics. The president of IMO lauded Deep Think's solutions, describing them as "astonishing in many respects."
