Google is giving its most advanced AI users something new to play with, and it’s called Deep Think. Starting today, users subscribed to Google AI Ultra can access Deep Think directly in the Gemini app. This new model is not just another iteration of Gemini 2.5; it’s based on the same variant that achieved top-tier scores at the International Mathematical Olympiad (IMO), a global competition where the best student minds go head-to-head on high-level maths problems. While that IMO version took hours to think through complex problems, the one in the Gemini app is tuned for speed and everyday usability — though still capable of serious reasoning, according to Google’s internal testing.
The focus with Deep Think is on better problem-solving. It uses what’s called parallel thinking, which means, instead of following one straight path to an answer, it considers multiple possibilities at once. This allows it to explore different ideas, make comparisons, and even revise its approach midway. Google says this makes it particularly good for creative or layered tasks — things like complex coding challenges, scientific research, and step-by-step design projects.
In a post announcing the release, Google said, “It’s a significant improvement over what was first announced at I/O,” referring to its earlier Gemini 2.5 demo. The model now reaches Bronze-level performance on the 2025 IMO benchmark while being far more responsive and usable in day-to-day tasks.
CEO Sundar Pichai also shared the news on X (formerly Twitter), writing: “We’re bringing a version of Deep Think that achieved gold-medal status at IMO to Ultra subscribers in the @Geminiapp (+ the official version is now in the hands of mathematicians). Toggle it on when reasoning through complex scientific literature, tackling a coding problem that requires careful consideration of time complexities – or anything else @DemisHassabis considers a fun Friday night.”
For those curious about how to use it, the feature is available through the Gemini app on mobile. Subscribers just need to enable the Deep Think toggle after selecting the 2.5 Pro model. At launch, users get a limited number of Deep Think prompts each day. Over time, Google plans to expand its availability, including access via the Gemini API for selected developers and academics.
Beyond performance, Google also says it’s worked on safety. The new model shows better content filtering and objectivity compared to previous versions, though it reportedly plays it safe and sometimes refuses even harmless queries.
– Ends