On May 6, Google introduced an updated version of its flagship generative AI — Gemini 2.5 Pro Preview (I/O edition). This version is available to users via the Gemini API, as well as on the Google AI Studio, Vertex AI platforms, and in the Gemini app for web and mobile devices. The update appeared ahead of the planned presentation at the annual Google I/O conference in response to positive feedback from the developer community.
Gemini 2.5 Pro Preview (I/O edition) has noticeably improved capabilities in programming, especially for creating interactive web applications. The model demonstrates increased accuracy in generating, editing, and transforming code, as well as in building complex “agent” workflows. According to the company, previously identified issues with function calls have been resolved, making the model more reliable.
In the WebDev Arena test, Gemini 2.5 Pro Preview (I/O edition) took first place, surpassing the previous version in key metrics, including the aesthetics and functionality of generated web applications. In the field of video understanding, the model achieved a score of eighty-four point eight percent on the popular VideoMME benchmark.
The updated model can be used both through developer interfaces and in the Gemini app, where it powers features such as “Canvas” for creating web applications using simple prompts. The price of Gemini 2.5 Pro Preview (I/O edition) remains unchanged compared to the previous version, allowing developers and a wide range of users to immediately try out the new capabilities.
