This is a sequel to the Gemini series first launched last December, and this time in the Gemini 2.5 Pro Experimental version – an experimental model that operates differently from the previous models. It focuses less quickly or efficiently, and more in the exercise of a thinking process that simulates reflection, analysis and synthesis of information.
Google emphasizes that it is a deep upgrade not only of the model base, but also of the post -learning training process. According to Cabokuzello readers, Google Depmind’s VP, the goal is now to “incorporate these thinking capabilities in each of the future models, so they can deal with more complex problems and support smart agents who are aware of context and task.”
As part of its impressive abilities, the model supports a huge amount of information-up to a million Tokans, which are about 750,000 words-and soon 2 million will be doubled. Meaning: You can enter extremely long texts, pictures, code, sound files and more – and he can still understand, process and analyze everything while maintaining consistency and focus.
In the field of development, Gemini 2.5 excels in working with code. It is able to write entire apps, edit and improve existing code, and even act as an autonomous agent that performs complex tasks for developers. Google is referred to as a “code monster”, which is especially suitable for engineers and professional developers.
Also in objective performance indices, the model has achieved particularly impressive results. Google notes that he broke records on several end tests, including the Humanity’s Last Exam test – a particularly challenging test that swears even the most skilled person.
The access to the model is already open to Gemini Advanced users through the GEMINI app on the phone, and also on the Google AI Studio platform through the browser. In the near future, access to Vertex AI users, Google’s cloud service, will also be expanded, enabling the model to combine the large -scale industrial and commercial tools.
Alongside this, Google continues to incorporate previous Gemini versions – like 2.0 – in its various products, including the renewed search engine with AI Mode and deep search tools Deep Research. In the new version, Google is expecting even smarter applications that can accompany users of decisions, surgery and depth studies.