mPLUG/GUI-Owl-1.5-32B-Think
mPLUG/GUI-Owl-1.5-32B-Think is a 33.4 billion parameter native GUI agent model, built on Qwen3-VL by X-PLUG, designed for multi-platform GUI automation across desktops, mobile devices, and browsers. This "Thinking" variant excels in complex tasks requiring planning and reflection, leveraging a scalable hybrid data flywheel and multi-platform environment RL. It achieves state-of-the-art performance on various GUI benchmarks, including OSWorld-Verified and WebArena, and supports long-horizon memory and external tool invocation.
Loading preview...
GUI-Owl 1.5-32B-Think: Advanced Multi-Platform GUI Agent
mPLUG/GUI-Owl-1.5-32B-Think is a 33.4 billion parameter model from the GUI-Owl 1.5 family, developed by X-PLUG and built upon Qwen3-VL. This "Thinking" variant is specifically designed for complex tasks that require advanced planning and reflection, distinguishing it from smaller "Instruct" models optimized for faster inference. It offers comprehensive multi-platform GUI automation capabilities, supporting desktops, mobile devices, and browsers.
Key Capabilities & Features
- State-of-the-Art Performance: Achieves top results across multiple GUI benchmarks, including OSWorld-Verified, AndroidWorld, OSWorld-MCP, Mobile-World, WindowsAA, WebArena, VisualWebArena, and WebVoyager.
- Tool & MCP Calling: Provides native support for invoking external tools and coordinating with MCP servers, demonstrating strong performance on OSWorld-MCP and Mobile-World.
- Long-Horizon Memory: Features built-in memory capabilities, outperforming other native agent models on MemGUI-Bench without requiring external workflow orchestration.
- Multi-Agent Ready: Can function as a standalone end-to-end agent or as specialized roles (planner, executor, verifier, notetaker) within the Mobile-Agent-v3.5 framework.
- Scalable Hybrid Data Flywheel: Leverages a unique data flywheel and multi-platform environment Reinforcement Learning (MRPO) for enhanced capabilities.
Ideal Use Cases
This model is particularly well-suited for developers and researchers building sophisticated GUI automation solutions, especially those involving complex, multi-step interactions across diverse operating systems and platforms. Its "Thinking" variant is optimized for scenarios demanding robust planning and reflective reasoning, making it a strong choice for challenging agentic tasks.