Google says its new “logical” models are the best post

After presenting a new “open” artificial intelligence model with Better performance on one graphics processing unitGoogle now has I made an update to the artificial intelligence models of its products using Gemini 2.5Which combines a “highly improved base model with improvement of training” to improve general performance. It is claimed that the first version, Gemini 2.5 Pro experimental, leads competition from Openaii, Anthropic, Xai, and Deepseek on common artificial intelligence standards that measure understanding, mathematics, coding and other capabilities. The new model is available to reach Google to study Or for Gemini subscribers in a drop -down application model menu.

The company also promotes the multiple media in Guemini as a feature, as it is able to explain not only the text, but also sound, fixed sound, video and symbol, and says that the window of a symbolic context of two million “will soon come” to help it process more data. Dimis Hasabis, CEO of Google DeepMind, Gemini 2.5 Pro is called a great model on the latest model, No. 1 on LMARNA through huge ELO points +39, with significant improvements in all areas of multimedia thinking, coding, and continuity, ” In a post on x.

Google says she jumped forward in quality because Gemini models are now “thinking” models that address step -by -step tasks and make more enlightened decisions, which they say lead to better answers and responses for complicated demands. Now, the blog post says, “… we are building these possibilities of thinking directly in all of our models, so that they can deal with the most complex problems and support for more capable agents.”

one Experimental video Show 2.5 Pro using these thinking possibilities for programming a single -wave video game: