Gemini 2.5 Pro Updated With Improved Coding Capabilities Ahead of Google I/O 2025

Google updated the existing Gemini 2.5 Pro artificial intelligence (AI) model with enhanced coding capabilities on Tuesday. The Mountain View-based tech giant is calling the newer version of the model Gemini 2.5 Pro Preview (I/O edition). Interestingly, the company said that it had planned to release this version at its upcoming Google I/O 2025 event, scheduled to be held between May  20-21. However, due to positive feedback for the model, the tech giant decided to release it earlier.

Gemini 2.5 Pro Gets New Coding Capabilities

In a blog post, the tech giant detailed the updated large language model. Google claims that with improved code understanding, the Gemini 2.5 Pro (I/O edition) can even build interactive web apps from scratch. Apart from user interface development, the model can also perform other complex coding tasks such as code transformation, code editing, and developing complex agentic workflows.

The updated model is now also ranking in the top position on the WebDev Arena leaderboard. The crowd-sourced leaderboard was developed by LMArena and measures human preference for a model's ability to complete web development tasks. The model previously ranked in second position, but has now taken the crown from Claude 3.7 Sonnet.

Gemini 2.5 Pro (I/O edition) also comes with a new video-to-code feature. The model's video understanding capability helped it score 84.8 percent on the VideoMME benchmark. Now, with improved coding performance, it can create an interactive web app based on a single YouTube video. This functionality is currently only available in Google AI Studio.

Additionally, the model is now also better at front-end web development. Google claims the AI model can be connected with an integrated development environment (IDE) and inspect components to match style properties such as colours, fonts, padding, margins, and more.

Developers can access and begin vibe coding with the updated Gemini 2.5 Pro model in the Gemini application programming interface (API) via Google AI Studio and Vertex AI. Regular users can find the new model in the Gemini app and web client. It also powers features such as Canvas.