Recently, StepFun officially released its first fully open-source GUI agent "GELab-Zero". This innovative product consists of two main parts: plug-and-play inference engineering infrastructure and a 4B GUI agent model that can run locally. GELab-Zero aims to provide users with an efficient and convenient agent experience.
The key feature of GELab-Zero is its lightweight local inference capability. This product supports running a 4B-scale model on consumer-grade hardware, ensuring low-latency response speed and effectively protecting user privacy. In addition, GELab-Zero is equipped with an一键 task start function, allowing users to complete automatic handling of environment dependencies and device management with just a single click, eliminating the cumbersome setup steps.
To meet diverse application needs, GELab-Zero provides multi-device task distribution functionality. Users can distribute tasks to multiple phones and can record interaction trajectories in real-time, facilitating subsequent observation and replication. More notably, this agent supports multiple working modes, including ReAct mode, Multi-Agent mode, and scheduled tasks, greatly enhancing its flexibility and adaptability.
In practical application scenarios, GELab-Zero has performed excellently. Official open-source benchmark tests conducted by the company show that GELab-Zero-4B-preview performs outstandingly in multiple dimensions such as GUI understanding, positioning, and interaction, especially in real mobile scenarios, demonstrating its strong application capabilities.
github:https://github.com/stepfun-ai/gelab-zero/
Key Points:
🌟 GELab-Zero is StepFun's first fully open-source GUI agent, supporting local deployment.
🚀 It has lightweight local inference capability, one-click start, and multi-device task distribution functions.
🏆 GELab-Zero has shown excellent performance in multiple benchmark tests and adapts well to real-world applications.
