Google unveiled its updated artificial intelligence model, Gemini 2.5 Pro Preview (I/O edition), ahead of its annual I/O developer conference. This enhanced version of their flagship AI offering reportedly delivers standout results across several mainstream performance benchmarks. It is now accessible through the Gemini API, as well as Google’s Vertex AI and AI Studio platforms, priced identically to the previous Gemini 2.5 Pro it replaces. Additionally, Google has integrated the Gemini 2.5 Pro Preview into its Gemini chatbot, available across both web and mobile platforms.
Google highlighted substantial enhancements, particularly in advanced programming tasks and interactive web application development. The updated model reportedly excels in areas like code transformation—modifying existing programs to achieve specific objectives—and precise code editing. According to company statements, Gemini 2.5 Pro Preview has been crafted in direct response to developer feedback, significantly reducing mistakes in API function calls and enhancing correct function triggers.
In recent leaderboard rankings, Gemini 2.5 Pro Preview attained the top position on the WebDev Arena Leaderboard, a metric that evaluates the capability of artificial intelligence models to build visually appealing and highly functional websites. It also achieved an impressive 84.8% result on VideoMME, a widely respected benchmark designed to measure video understanding performance, further underscoring advancements beyond just coding.
Google described the model as uniquely adept at creating aesthetically advanced web solutions while still effectively responding to developer guidance. The launch occurs amid intense competition in the AI sector, with rivals OpenAI and xAI expected to soon reveal their respective state-of-the-art AI products. At their upcoming I/O event, Google is expected to announce further updates and expand its portfolio of AI-based technologies and tools.
