On Tuesday, Google announced the launch of Gemini 2. 5 Pro Preview (I/O edition), an updated iteration of its flagship AI model, Gemini 2. 5 Pro, which the company claims outperforms several widely-used benchmarks. Gemini 2. 5 Pro Preview (I/O edition) is accessible through the Gemini API as well as Google’s Vertex AI and AI Studio platforms, and is priced the same as the Gemini 2. 5 Pro model it replaces. Additionally, it is integrated into Google’s Gemini chatbot app, available on both web and mobile devices. The release of this model precedes Google’s annual I/O developer conference—hence the “I/O edition” label—where the company is expected to introduce numerous models alongside AI-powered tools and platforms. Google is actively vying for attention and market share in the highly competitive AI landscape, with rivals such as OpenAI and xAI poised to launch models anticipated to deliver robust performance. Google states that Gemini 2. 5 Pro Preview (I/O edition) features “significantly” enhanced capabilities in coding and developing interactive web applications.

The model also demonstrates improvements in tasks like code transformation—modifying code to meet specific objectives—and code editing. In a blog post, Google highlights that Gemini 2. 5 Pro Preview (I/O edition) leads the WebDev Arena Leaderboard, a standard benchmark assessing a model’s ability to create visually appealing and functional web apps. Furthermore, the model achieves state-of-the-art results in video understanding, scoring 84. 8% on the popular VideoMME benchmark. “For developers already using Gemini 2. 5 Pro, this new version will not only boost coding performance but also address key developer feedback by reducing errors in function calling and enhancing function calling trigger rates, ” Google stated in its blog. “By default, the model favors aesthetic web development while maintaining its steerability. ”