Gemini 2.5 Pro Preview delivers enhanced coding capabilities ahead of Google I/O, focusing on front-end development and complex code transformations. Google DeepMind reports improved performance in agentic workflows and reduced errors in function calling compared to prior versions. The model powers tools like Cursor's code agent and Replit's platforms, enabling more reliable coding assistance.

The updated model leads web development on the WebDev Arena leaderboard, excelling at creating functional and visually appealing web applications. Developers can now implement features like adding styled video players directly in IDEs, with the model understanding design properties such as colors and spacing. This advancement simplifies translating concepts into working apps with polished UIs, as demonstrated by the dictation starter app featuring animated microphone UIs.

Gemini 2.5 Pro also achieves 84.8% on VideoMME benchmarks, enabling new applications like converting YouTube videos into interactive learning tools. Technical evaluations show it handles complex tasks requiring senior-level judgment, such as refactoring backend request routing systems. Enterprise access remains via Vertex AI, with the model available at unchanged pricing.

Developers can access Gemini 2.5 Pro through Google AI Studio. The model card update reflects version 05-06, automatically serving improvements to existing users. Reduced latency combined with high capability makes it suitable for real-time coding tasks requiring both speed and accuracy.