Google has rolled out an upgraded version of its Gemini 2.5 Pro AI model ahead of the upcoming Google I/O 2025 event. Originally planned for release during the May 20-21 developer conference, the enhanced model – now called Gemini 2.5 Pro Preview (I/O edition) – is out early due to strong feedback.
This new version focuses heavily on improving coding tasks. Google says the AI can now create full web apps from scratch, edit existing code, transform code, and even manage complex workflows that mimic agent-like behavior. It also now leads the WebDev Arena leaderboard, a platform run by LMArena that tracks AI performance in web development. Previously second, Gemini 2.5 Pro has now overtaken Claude 3.7 Sonnet to reach the top.
A standout new feature is its ability to convert video into code. Using video understanding, the model scored an impressive 84.8% on the VideoMME benchmark. This means it can take a YouTube video and turn it into a working interactive web application – a feature currently supported in Google AI Studio.
Front – end development has also seen upgrades. The model can connect with IDEs and examine style components like fonts, colors, spacing and margins to match visual designs more accurately.
Developers can try the updated model through the Gemini API via Google AI Studio and Vertex AI. Regular users will find it in the Gemini App and web version. The model also supports tools like Canvas, offering more powerful features for everyday users and codes alike.
By
