Google on Tuesday introduced early access to Gemini 2.5 Pro Preview (I/O edition), an updated version of its Gemini 2.5 Pro model designed to improve coding tasks, particularly for creating interactive web applications.
Tulsee Doshi, Senior Director of Product Management, said the update was initially planned for release at Google I/O in a few weeks. However, due to strong interest in the model, Google decided to make it available sooner to allow developers to begin building.
Enhanced Features of Gemini 2.5 Pro
Doshi noted that the update builds on positive feedback about Gemini 2.5 Pro’s coding and multimodal reasoning abilities.
The model now offers improved support for:
- Code transformation
- Code editing
- Developing complex agentic workflows
- User interface development
The updated model leads the WebDev Arena Leaderboard, outperforming its predecessor by 147 Elo points. This leaderboard evaluates a model’s ability to create visually appealing and functional web applications based on human preferences.
The model also maintains strong performance in native multimodality and long-context processing, achieving an 84.8% score on the VideoMME benchmark for video understanding.
Availability and Access
Developers can use the updated Gemini 2.5 Pro via the Gemini API on Google AI Studio and Vertex AI, while the Gemini app incorporates the model to support Canvas, allowing users to create code and build interactive web applications with a single prompt.