Better than GPT o3 and Grok 4: Google Releases Gemini 2.5 Deep Think

Credit: Google

Google’s DeepMind division has announced the launch of Gemini 2.5 Deep Think, the company’s most advanced AI model, according to its developers. It can respond to user queries by exploring and considering multiple options simultaneously, then comparing the results and choosing the best one. Owners of Google’s $250-per-month Ultra subscription will get access to the AI model this week.

Deep Think mode is based on the Gemini 2.5 Pro model, but has a number of its own features. Firstly, it has an increased computational resource for reasoning, which is important for tasks that require a long solution. Secondly, Deep Think mode includes several AI agents that, working in parallel, try different approaches to solving the problem. As a result, the critic module either chooses the best solution or combines the final answer from several solutions.

Gemini 2.5 Deep Think automatically deploys the right AI agents, for example, to write code or search the web. According to the developers, the neural network is capable of producing “much longer answers” compared to traditional AI models. Google plans to open access to Gemini 2.5 Deep Think to a select group of developers via the corresponding API in the near future.