mPLUG/GUI-Owl-32B

VISIONPricing:Input $1.06 / Cached $0.053 / Output $2.6Concurrent Unit Cost:2Model Size:32BQuant:FP8Context Size:32kPublished:Aug 15, 2025License:mitArchitecture:Transformer0.0K Open Weights Featherless Exclusive Cold

GUI-Owl-32B is a 32 billion parameter model developed by mPLUG as part of the Mobile-Agent-V3 project, designed for GUI automation tasks. It achieves state-of-the-art performance across various GUI benchmarks, including ScreenSpot-V2, ScreenSpot-Pro, OSWorld-G, MMBench-GUI, Android Control, Android World, and OSWorld. With a context length of 32768 tokens, this model excels at automating complex interactions within graphical user interfaces and can be integrated into multi-agent frameworks for advanced applications.

Loading preview...

GUI-Owl-32B Overview

GUI-Owl-32B is a 32 billion parameter model developed by mPLUG, specifically engineered for advanced Graphical User Interface (GUI) automation. It is a core component of the Mobile-Agent-V3 project, focusing on enabling intelligent agents to interact with and control GUIs across various platforms.

Key Capabilities

  • State-of-the-Art GUI Automation: Achieves leading performance on a wide array of GUI automation benchmarks, including ScreenSpot-V2, ScreenSpot-Pro, OSWorld-G, MMBench-GUI (L1, L2), Android Control, Android World, and OSWorld-Verified.
  • Multi-Agent Integration: Designed to be instantiated as specialized agents within the Mobile-Agent-V3 multi-agent framework, facilitating the accomplishment of more complex, multi-step tasks.
  • High Context Length: Supports a maximum model length of 32768 tokens, allowing for processing extensive GUI states and interaction histories.
  • Flexible Image Input: Configurable to receive multiple images per prompt, with adjustable pixel limits, enabling detailed visual understanding of GUI elements.

Good For

  • Automating complex GUI workflows: Ideal for tasks requiring precise interaction with desktop or mobile application interfaces.
  • Developing intelligent agents for mobile and desktop environments: Provides a robust foundation for creating agents that can navigate and operate GUIs autonomously.
  • Research in GUI understanding and control: Offers a strong baseline and framework for advancing the field of GUI automation and agent-based interaction.
  • Applications requiring high-performance GUI benchmark results: Demonstrates strong performance across established benchmarks, indicating reliability for demanding use cases.

Finetunes
1 models