- Google Gemini 2.5 Pro is obtaining a new deep thinking model
- Deep think allows Gemini to consider multiple reasoning routes before answering
- Propunde thinking will improve the accuracy of Gemini in Mathematics and Complex Code
Google is adding a little additional intellectual capacity to Gemini with a new deep way of thought. The company presented the last option for Google Gemini 2.5 Pro on Google I/O 2025, showing what AI can do with additional depth.
Profindes to think basically increases the ‘mind’ of Gemini with additional brains. Gemini in deep way of thinking will not only spit an answer to a consultation as quickly as possible. Instead, execute multiple possible lines of parallel reasoning before deciding how to respond. It is like the equivalent of looking in both directions, or rereading the instructions before building a furniture.
And if Google tests are something to happen, Deep Think’s Brainpower is working. He is working at a first level level in the US Mathematics Olympiad. Profunda thinks that it is not yet widely available. Google is throwing it to trusted testers just for now. But, presumably, once all springs are resolved, everyone will have access to the deepest thoughts of Gemini.
Gemini shines
Deep Think is directly adjusted to the rest of the growing Gemini 2.5 line and the new features that arrive for their various models in the API used by developers to embed Gemini in their software.
For example, Gemini 2.5 Pro now admits the generation of native audio. That means you can talk to you. The speech has a characteristic of “affective dialogue”, which detects emotional changes in its tone and fits accordingly. If it sounds stressed, Gemini could stop speaking as a patient’s customer service agent and respond more as an empathetic and reflective friend (or at least how AI interprets such an answer). And it will be better to know when to speak at all thanks to the new proactive audio function, which filters the background noise so that Gemini only intervenes when he is sure he is talking to her.
Together with new security safeguards and the next computer use functions of the Mariner project, Gemini 2.5 is trying to be the AI in which he trusts not only with his calendar or code, but with the narration of his book or the entire operating system.
Another element that expands in Gemini 2.5 is what Google calls a “thought budget.” Previously exclusive Gemini 2.5 Flash, the thought budget allows developers to decide how deep the model should think before responding. It is a good way to make sure you get a complete answer without spending too much. Otherwise, deep thinking could give him a sample of his reasoning, or give everything and make it too expensive for any follow -up.
In case it is not clear to what these thoughts imply, Gemini 2.5 Pro and Flash will offer ‘thought summaries’ for developers, a document that shows the exact details of what the AI was doing in terms of applying information through their reasoning process, so it can look inside the AI brain.
All this points to a pivot of the models that only speak quickly with emphasizing those that can reason deeper, although slower. Deep thinking is part of that change towards deliberate reasoning and in layers. It is no longer just trying to predict the next word, but applies that logic to ideas and the process of obtaining answers to your questions. Google seems interested in making Gemini not only can you get answers, but to understand the form of the question itself.
Of course, the reasoning of AI still exists in a space where a perfectly logical response could come with a random side of nonsense, regardless of how impressive the reference score is. But you can begin to see the form of what is coming, where the promise of a real “co -pilot” comes to fruition.
