Google has introduced the Gemini 2.5 Computer Use AI, a model specifically designed to interact with user interfaces (UIs). Built on the advanced Gemini 2.5 Pro framework, this AI leverages its visual and reasoning strengths to drive AI agents. It is capable of navigating both web browsers and Android UI environments.
According to Google, the Gemini 2.5 Computer Use AI can perform human-like actions such as clicking, typing, and scrolling to complete tasks. On the WebVoyager benchmark, it achieved an impressive 88.9%, slightly surpassing OpenAI’s Computer-Using AI Agent, which scored 87%. Similarly, in the Online-Mind2Web benchmark, Google’s model outperformed OpenAI’s Operator AI.
This demonstrates Google’s success in developing a top-tier AI model that can execute tasks reliably on browser platforms, showing advantages over competitors like Claude Sonnet 4.5 and OpenAI’s AI agents in both speed and accuracy.
The Gemini 2.5 Computer Use AI is already integrated into Google initiatives such as Project Mariner and AI Mode in Google Search. Additionally, developers can access the model via the Google AI Studio and Vertex AI APIs.
