Google reveals its advanced reasoning feature, Gemini 2.5 Deep Think, for its AI ultra subscribers. The tech major claimed it as the most advanced reasoning model released yet and expected to give tough competition to rivals, including Grok-4.
According to reports, Google tested Deep Think on multiple AI benchmarks, including LiveCodeBench, and it outpaced competitors such as xAI’s Grok-4. Moreover, only AI Ultra Subscribers can leverage the Deep Think feature from today. Although, for now, the number of prompts allowed is limited.
Google wrote in their blog about the release, ” We continue to build safety and responsibility into Gemini throughout the training and development lifecycle. Gemini 2.5 Deep Think demonstrated improved content safety and tone objectivity compared to the Gemini 2.5 Pro, but did have a higher capacity tendency to refuse benign requests.”
“We are taking a closer look, including frontier safety evaluations and the implementation of planned mitigations for critical capability levels,” the blog added.
Deep Think uses the ‘Parallel Thinking’ techniques, offering great results to users. This latest approach, integrated by Google, enables Gemini to generate multiple ideas at once and consider all of them simultaneously. Moreover, this is also a multi-modal technology, so users can give prompts in different forms, including text, images, data, or sound.
