- Google Gemini 2.5 Pro gets a new deep -minded model
- Deep Think allows Gemini to consider more reasoning paths before responding
- Deep Think will improve Gemini’s accuracy on complex math and code
Google adds some extra brain power to Gemini with a new deep -minded state. The company revealed the latest option for Google Gemini 2.5 Pro on Google I/O 2025, which shows what it is AI can do with extra depth.
Depth thinking basically think Geminis ai ‘Mind’ with additional brains. Gemini in deep state of thinking doesn’t just spit an answer to an inquiry as soon as possible. Instead, it runs several possible lines of reasoning parallel before deciding how to respond. It is like the AI equivalent to see both ways or re -read the instructions before building a piece of furniture.
And if Google’s tests are something to pass by, Deep Think’s brain power works. It appears at a top level level of 2025 US Math Olympiad, coming out on top in Live Codebench-competitive programming test, scoring an incredibly high 84% on the popular MMMU, a kind of Decathlon of multimodal reasoning tasks. Deep Think is not widely available yet. Google is rolling out to trusted testers only for now. But presumably, when all the kinks are ironed out, everyone will have access to the deepest of Gemini’s thoughts.
Gemini shines on
Deep Think fits right into the rest of the Gemini 2.5’s growing lineup and the new features arriving at its different models in API used by developers to integrate Gemini into their software.
E.g. Supports Gemini 2.5 Pro now native sound generation. This means it can talk back to you. The speech has an “affective dialog” feature that detects emotional shifts in your tone and adjusts accordingly. If you sound stressed, Gemini may stop talking like a patient customer service and react more like an empathetic and thoughtful friend (or at least how AI interprets such an answer). And it will be better at knowing when to talk at all thanks to the new proactive sound feature that filters background noise, so Gemini is only belled in when it’s sure you’re talking to it.
The paired with new security protection measures and the upcoming Project Mariner Computer Features Try Gemini 2.5 very hard to be AI, you trust not only with your calendar or code but with your book narrative or the entire operating system.
Another element expanded over Gemini 2.5 is what Google calls a ‘thinking budget’. Previously unique for Gemini 2.5 -flash, the thinking budget lets developers decide how deep the model should think before responding. It’s a good way to make sure you get a full answer without using too much. Otherwise, deep thinking could give you just a taste of its reasoning, or give you it all and make it too expensive for any follow -up.
In the event that it is not clear what these thoughts are involved, Gemini 2.5 Pro and Flash will offer ‘thought overviews’ for developers, a document that shows the exact details of what AI did with regard to the use of information through its reasoning process so that you can actually see inside the AI brain.
All this signalizes a turn from models that just speak quickly to emphasize those that can resonate deeper if it is slower. Deep Think is part of this shift against deliberate, layered reasoning. It’s not just trying to predict the next word anymore, it applies this logic to ideas and the very process of providing answers to your questions. Google seems eager to make Gemini not only able to get answers, but to understand the shape of the question itself.
Of course, AI resonance still exists in a room where a perfect logical answer may come with a random side of rubbish, no matter how impressive benchmark scores. But you can start to see the shape of what comes, where the promise of an actual ‘co-pilot’ AI is going to perform.
