Just days before the annual Google I/O conference, Google announced an updated version of its flagship programming model, named Gemini 2.5 Pro Preview (I/O release).
The model is now available through the Gemini API, Vertex AI, and AI Studio interfaces. Furthermore, its scope is being expanded via the Gemini app for web and mobile devices.
![Screenshot showing the Gemini models selection list in Google AI Studio, highlighting the Gemini 2.5 Pro Preview 05-06 model marked [NEW]](https://aiarabai.com/wp-content/uploads/2025/05/Gemini-2.5-Pro-Preview-ai-studio.jpg "Gemini models list available in Google AI Studio")
The company stated that the new version will replace the previous iteration at no change in price, with official adoption within the Gemini app on the web and mobile devices.
This move comes at a strategic time, as major companies like OpenAI and xAI are preparing to launch their anticipated models.
Clearly, Google aims to solidify its position in this race, especially with its emphasis that Gemini 2.5 Pro Preview demonstrates advanced performance across several benchmarks, most notably in the areas of interface development and video understanding.
Impressive Capabilities in Web Application Creation
According to what Google published on its official blog, the new model boasts improved capabilities in code editing, alongside a better understanding of interactive application structure.
It has also secured the first place in the WebDev Arena ranking, a benchmark that measures how well models create elegant and practical web applications from a human user's perspective.
Furthermore, the model achieved a score of 84.8% in the VideoMME test, positioning it among the best in the field of video analysis and component understanding.
Transforming YouTube Videos into Interactive Applications
Among the notable use cases showcased by Google is Gemini 2.5 Pro's ability to transform an educational YouTube video into an interactive learning application. This is achieved through a template developed within Google AI Studio.
The model can handle interfaces, colors, edges, subtle movements, and design features like a visually interactive microphone button.
Convert YouTube clips to learning applications now here
Michel Cadasta, Head of Replit, indicated that Gemini 2.5 Pro is considered the best in terms of balancing performance and speed, confirming it will be adopted for tasks requiring low latency with high execution time.
Meanwhile, the Cognition team commented that the new model was the first of its kind capable of solving a development test requiring extensive restructuring of a request routing system. This made it appear like a seasoned developer in terms of judging architectures and selecting abstractions.
More Effective with Fewer Errors
A notable aspect of this release is its responsiveness to developer feedback. Consequently, Google has improved programming function calls and reduced error rates in these operations.
Furthermore, the model has become more steerable, which makes its integration into advanced development environments more effective.
Google states that current developers using the 2.5 Pro version will not need to take any action, as the previous release (03-25) automatically points to the newer version (05-06).
This model's capabilities are expected to be extensively demonstrated during Google I/O events. This reflects a desire to highlight programming innovations rather than solely focusing on showcasing the general abilities of the models.
With this update, Google has laid a new cornerstone for competition in the areas of web development and converting videos into practical applications.
It also confirms once again that Gemini models are not just text generators, but true development tools.
