GELab-Zero - Open source end-side multimodal GUI Agent model by Steps team

What is GELab-Zero?

GELab-Zero is an open source end-side multimodal GUI Agent model built on the Qwen3-VL-4B-Instruct base model with 4B parameters, which recognizes UI elements and performs clicks, swipes, etc., supports cross-application tasking (e.g., takeout, travel, etc.), and has the ability of zero sample adaptation to adapt to unseen apps. The model is open-sourced using the Apache 2.0 protocol, supports Ollama fast startup, automatically handles ADB connections and dependency installation, and provides task recording playback. In the AndroidDaily benchmark test, the accuracy rate reaches 73.4%, the performance exceeds the mainstream models of the same size, and outperforms GUI-Owl-32B which has a larger number of parameters.

Features of GELab-Zero

• Local Deployment and Privacy: Supports local operation without relying on the cloud, ensuring data privacy and low-latency operations.
• Lightweight design: Optimized 4B models run efficiently on consumer-grade hardware, balancing performance and resource consumption.
• One-Click Deployment: Provides a complete deployment process that automates environmental dependencies and device management, simplifying the threshold for use.
• Multi-device support: Supports multi-device connectivity and task distribution, facilitating task operations on different devices.
• multimodal interaction: Supports a variety of interaction modes, such as ReAct Closed-loop, multi-intelligence body collaboration and timed tasks to adapt to complex scenarios.
• Dynamic task scheduling: Supports distributed execution of tasks and interactive trajectory logging for easy task management and reproduction.
• Generic GUI Understanding: Recognizes and operates a variety of mobile application interfaces without the need for adaptation by application developers.
• Enterprise Application Support: Business users can directly reuse the infrastructure and quickly integrate it into their product operations.
• Open Source and Scalability: Provide open source code and infrastructure to support customization and extensions by developers.

Core Benefits of GELab-Zero

• Privacy and Local Deployment: Supports local operation without relying on the cloud, ensuring data privacy and low-latency operations.
• Lightweight & High Performance: 4B model optimized to run efficiently on consumer-grade hardware, balancing performance and resource consumption.
• One-Click Deployment Experience: Provides a complete deployment process that automates environmental dependencies and device management, simplifying the threshold for use.
• Multi-device and multi-tasking support: Supports multi-device connectivity and task distribution, which facilitates task operations on different devices and improves efficiency.
• Multimodal interaction capabilitiesIt supports a variety of interaction modes, such as ReAct closed-loop, multi-intelligence collaboration and timed tasks, to adapt to the needs of complex scenarios.
• Generic GUI Understanding: Recognizes and operates a variety of mobile application interfaces without the need for application developers to adapt, and has wide versatility.
• Enterprise Application Integration: Business users can directly reuse infrastructure to quickly integrate GUI Agent capabilities into their product operations.
• Open Source and Scalability: Provide open source code and infrastructure to support developers to customize and extend and facilitate technology iteration.
• High Performance Benchmark Performance: Excellent performance in several benchmarks, especially leading accuracy in the AndroidDaily benchmark, validating its strong task execution capabilities.

What is GELab-Zero's official website

• Project website:: https://opengelab.github.io/
• Github repository:: https://github.com/stepfun-ai/gelab-zero
• HuggingFace Model Library:: https://huggingface.co/stepfun-ai/GELab-Zero-4B-preview

Individuals for whom GELab-Zero is indicated

• developers: Developers who want to rapidly deploy and use GUI Agent can customize and extend it with open source code and infrastructure.
• business user: Organizations that need to integrate GUI Agent capabilities into their product operations can directly reuse GELab-Zero's infrastructure to quickly implement functionality.
• research worker: Scholars and researchers working in the fields of artificial intelligence, automated interaction, etc., can use models and benchmarking for research and innovation.
• Mobile Application Developers: Developers who want to integrate automated interactions into their mobile apps can use GELab-Zero's universal GUI comprehension capabilities without additional adaptations.
• technology enthusiast: Individual users interested in the GUI Agent and automated task execution can experience its capabilities through local deployment.
• educator: Teachers and educational institutions that need automated support tools in the field of education can use GELab-Zero to assist in teaching and learning tasks.