Google has decided to anticipate the release of the updated version of Gemini 2.5 Pro, originally scheduled for the imminent annual conference dedicated to the Google I/O developers. The new version, called Gemini 2.5 Pro Preview (I/O Edition)is now available in early access and brings significant improvements in the skills of codingwith particular focus on the development of web interactive apps. Let’s see all the details and news of this new model together.

The new Gemini 2.5 Pro Preview (I/O Edition) model is particularly capable in web development

The decision to accelerate the launch was motivated by the enthusiasm generated by the previous versions of the model. As explained by Tulsee DoshiSenior Director of Product Management of Google, the company preferred to make these new features available to developers and users immediately, without waiting for the developer conference to be held in a couple of weeks.

The new Gemini 2.5 Pro stands out mainly for its enhanced skills in the field of web development, but the update goes beyond the simple creation of user interfaces. The improvements extend to numerous aspects of programmingwhich include, among other things, the transformation of the code, advanced editing and the development of “agentic” complex complex workflows (it is, for those who do not chew artificial intelligence, systems capable of operating in a semi-autonomous way).

The effectiveness of these improvements is supported by concrete results: The updated model conquered the first position in the WebDev Arena Leaderboard, overcoming the previous version with a positive margin of +147 points Elo. For those who do not know it, this ranking measures human preference towards the skills of an artificial intelligence model in creating functional and aesthetically pleasing web apps.

However, it is not a monothematic improvement: the new Gemini 2.5 Pro keeps e strengthens its excellent multimodal skills and management of long contexts, reaching respectable performance in video understanding with a score of 84.8% in the Benchmark Videomme.

https://www.youtube.com/watch?v=6lnusuyv6rq

The goodness of this new model was also confirmed by users who have had the opportunity to try the skills of Gemini 2.5 Pro Preview (I/O Edition) in the professional field.

Michael TruellCEO of Cursor, one of the most famous and widely used integrated development environments based on artificial intelligence, thus comments on the release of this new model:

We are enthusiastic about the latest version of Gemini 2.5 Pro, which is based on its already solid programming skills in the real world. We are internally observing that the new model presents a significant reduction in tool call problems, an improvement that we believe our users will find useful and that will make the 2.5 pro version even more effective than first in cursor.

There availability of the updated model is immediate. The developers can begin to use it through the Gemini API via Google to the study and vertex AI, while end users can access it directly through the Gemini app, where it feeds features such as Canvas (introduced a few weeks ago together an audio Ovserview) and allows you to create web interactive apps with a single prompt.