Google introduces Gemini 2.5 Pro I/O Edition with significantly improved programming performance
Google recently introduced a new version of its AI model called Gemini 2.5 Pro Preview (I/O Edition), which brings significant improvements in programming and application development. Released ahead of the Google I/O conference, this update represents an important step forward in the field of AI assistants for developers.
Improved front-end and UI development
Gemini 2.5 Pro I/O Edition excels particularly in the area of user interface and web application development. The model ranked first on the WebDev Arena leaderboard, which evaluates human preferences for the quality of web applications created using AI. Developers can therefore use Gemini to create aesthetically appealing and functional web applications with much better results than before. The model now has a strong sense of aesthetic design and can quickly transform concepts into working applications with visually attractive elements, such as responsive layouts or animated UI components. As an example, Google highlights the microphone animation in a sample dictation application.
Code transformation and editing
One of the new model’s greatest strengths is its ability to transform existing code, edit it efficiently, and support sophisticated agentic workflows. Gemini 2.5 Pro now handles tasks such as code refactoring better and can make sound judgments and choices about abstractions – much like a senior developer during complex refactoring. As Michele Catasta, president of Replit, states: "We found that Gemini 2.5 Pro is the best frontier model when it comes to the 'capability versus latency' ratio... It was able to make sound judgments and choose good abstractions."
Another significant improvement is the reduction of errors related to function calling and an improved function call execution rate, making integrations more reliable. This improvement directly responds to feedback from developers and addresses one of the common problems of previous models.
Results on key benchmarks
Gemini 2.5 Pro I/O Edition achieves excellent results on several important performance benchmarks:
- First place on the WebDev Arena leaderboard for web application development.
- Excellent results on evaluations for junior developers, including solving complex backend refactoring tasks that other models previously could not solve.
- High score on SWE-Bench Verified (the industry standard for evaluating agentic coding), demonstrating robust performance across real-world software engineering scenarios.
Availability and deployment
Developers can access the updated model through:
- Google AI Studio using the Gemini API
- Vertex AI (for enterprise customers)
No additional steps are required from existing Gemini 2.5 Pro users – all endpoints now point to the latest version at no additional cost. The updated model card reflects these improvements as of May 6.
