A new family of vision-language models (VLMs) called Holo1 has been introduced, specifically designed for graphical user interface (GUI) automation. These models power the recently released GUI agent, Surfer-H, which aims to streamline interactions with digital interfaces.
Holo1 models are trained to understand and navigate on-screen elements, enabling Surfer-H to perform tasks such as clicking buttons, filling forms, and extracting information from web pages or desktop applications. The agent leverages the Holo1 family's visual understanding to interpret screenshots and execute actions based on natural language instructions.
According to the developers, Holo1 achieves state-of-the-art performance on several GUI automation benchmarks, outperforming existing models in accuracy and speed. The models are available in different sizes to accommodate various deployment scenarios, from edge devices to cloud servers.
Surfer-H, built on top of Holo1, can be integrated into workflows for web scraping, UI testing, and robotic process automation (RPA). Early adopters report significant reductions in manual effort for repetitive tasks.