Google Introduces New "Deep Think" Feature for Gemini 2.5 Pro Model

Google has unveiled a brand new pondering mode referred to as Deep Suppose, added to its Gemini 2.5 Professional mannequin. The corporate introduced that this newly added function achieved vital success in testing. The Deep Suppose function is at the moment accessible through the Gemini API, although it’s restricted to pick customers for now.

Due to this innovation, the AI can now consider a number of prospects earlier than answering a query, resulting in extra correct outcomes. In keeping with Google, Gemini 2.5 Professional with Deep Suppose achieved the very best rating in LiveCodeBench, one of the difficult coding benchmarks. It additionally outperformed OpenAI’s o3 mannequin within the MMMU check, which measures notion and reasoning skills.

At present, solely trusted check customers can entry Deep Suppose by means of the Gemini API. Google hasn’t shared detailed details about how the function works however famous that it’s based mostly on parallel pondering methods. The corporate additionally said that further safety testing will likely be carried out earlier than the function is made extensively obtainable.

Along with Deep Suppose, Google introduced an replace for the Gemini 2.5 Flash mannequin. This up to date model brings enhancements in coding, multimedia processing, lengthy content material comprehension, and logical reasoning. Google shared that the up to date 2.5 Flash mannequin is now obtainable for trial use through AI Studio, Vertex AI, and Gemini apps, and will probably be usually obtainable to builders in June.

Google additionally launched a brand new AI mannequin referred to as Gemini Diffusion, claiming it operates 4 to five occasions quicker than comparable programs whereas providing comparable efficiency to bigger fashions. Gemini Diffusion can be at the moment obtainable solely to trusted testers.