nielsr's picture
nielsr HF Staff
Improve model card: add metadata and links to paper/code
9a838e7 verified
|
raw
history blame
2.4 kB
metadata
language:
  - en
license: mit
pipeline_tag: image-text-to-text
library_name: transformers

Mobile-Agent-v3.5 (GUI-Owl-1.5)

This repository contains the model weights for Mobile-Agent-v3.5 (also known as GUI-Owl-1.5), a native multi-platform GUI agent foundation model.

Introduction

GUI-Owl-1.5 is a native multi-platform GUI agent model family featuring instruct and thinking variants. It supports a wide range of platforms, including desktop, mobile, and browser environments, to enable cloud-edge collaboration and real-time interaction. The model unifies perception, grounding, reasoning, planning, and action execution within a single policy network.

Key features include:

  • Hybrid Data Flywheel: A data pipeline for UI understanding and trajectory generation based on simulated and cloud-based sandbox environments.
  • Unified Enhancement of Agent Capabilities: A unified thought-synthesis pipeline to enhance reasoning, memory, and Tool/MCP (Model Context Protocol) usage.
  • Multi-platform Environment RL Scaling: Uses a new environment RL algorithm, MRPO, to address challenges in multi-platform interaction and long-horizon task training.

Benchmarks

GUI-Owl-1.5 achieves state-of-the-art results on more than 20 GUI benchmarks:

  • GUI Automation: OSWorld (56.5), AndroidWorld (71.6), and WebArena (48.4).
  • Grounding: ScreenSpotPro (80.3).
  • Tool-calling: OSWorld-MCP (47.6) and MobileWorld (46.8).
  • Memory & Knowledge: GUI-Knowledge Bench (75.5).

Citation

If you find this model useful, please cite our paper:

@article{MobileAgentv3.5,
  title={Mobile-Agent-v3.5: Multi-platform Fundamental GUI Agents},
  author={Haiyang Xu, Xi Zhang, Haowei Liu, Junyang Wang, Zhaozai Zhu, Shengjie Zhou, Xuhao Hu, Feiyu Gao, Junjie Cao, Zihua Wang, Zhiyuan Chen, Jitong Liao, Qi Zheng, Jiahui Zeng, Ze Xu, Shuai Bai, Junyang Lin, Jingren Zhou, Ming Yan},
  journal={arXiv preprint arXiv:2602.16855},
  year={2026}
}