$git clone https://github.com/Tongyi-MAI/MAI-UIWe present MAI-UI, a family of GUI agent foundation models spanning the full spectrum of sizes, including 2B, 8B, 32B, and 235B-A22B variants. Our core contribution includes: - 🔧 Agent-user interaction and MCP augmentation: enabling agent to interact with
| 1 | <p align="center"> |
| 2 | <img src="./assets/img/mai_title_v1.png" alt="MAI-UI" /> |
| 3 | </p> |
| 4 | |
| 5 | <p align="center"> |
| 6 | <a href="https://arxiv.org/abs/2512.22047"><img src="https://img.shields.io/badge/📄%20arXiv-Paper-red" alt="arXiv" /></a> |
| 7 | <a href="https://tongyi-mai.github.io/MAI-UI-blog/"><img src="https://img.shields.io/badge/🌐%20Website-MAI%20UI-blue" alt="Website" /></a> |
| 8 | <a href="https://tongyi-mai.github.io/MobileWorld/"><img src="https://img.shields.io/badge/🌐%20Website-MobileWorld-blue" alt="Website" /></a> |
| 9 | <a href="https://huggingface.co/Tongyi-MAI"><img src="https://img.shields.io/badge/🤗%20Hugging%20Face-Model-orange" alt="Hugging Face Model" /></a> |
| 10 | <a href="https://www.modelscope.cn/organization/Tongyi-MAI"><img src="https://img.shields.io/badge/🤖%20ModelScope-Model-purple" alt="ModelScope Model" /></a> |
| 11 | </p> |
| 12 | |
| 13 | We present **MAI-UI**, a family of GUI agent foundation models spanning the full spectrum of sizes, including **2B**, **8B**, **32B**, and **235B-A22B** variants. Our core contribution includes: |
| 14 | - **🔧 Agent-user interaction and MCP augmentation**: enabling agent to interact with user and use MCP tools to complete the task. |
| 15 | - **☁️ Device–cloud collaboration system**: dynamically selecting on-device or cloud execution based on task execution state and data sensitivity. |
| 16 | - **📈 Dynamic RL Scaling**: large-scale reinforcement learning with scaling parallel environments (up to **512**) and context length (up to **50**). |
| 17 | - **🏆 State-of-the-Art Performance**: MAI-UI establishes new benchmark SOTA results across GUI grounding and navigation tasks. |
| 18 | |
| 19 | <p align="center"> |
| 20 | <img src="./assets/img/overview.png" alt="Overview" width="800"> |
| 21 | <br> |
| 22 | <em>Overview of MAI-UI performance</em> |
| 23 | </p> |
| 24 | |
| 25 | ## 📰 News |
| 26 | * **[2026-03-20]** 📄 **Blog Posts**: Our [Grounding](https://tongyi-mai.github.io/MAI-UI-blog/Grounding-Blog) and [Navigation](https://tongyi-mai.github.io/MAI-UI-blog/MobileWorld-Blog-Post) Blog Posts are available now! |
| 27 | * **[2026-01-15]** 🥇 **New Record on AndroidWorld**: MAI-UI-235B takes #1 on the [AndroidWorld Leaderboard](https://docs.google.com/spreadsheets/d/1cchzP9dlTZ3WXQTfYNhh3avxoLipqHN75v1Tb86uhHo/edit?gid=0#gid=0) for pure-vision, end-to-end models with a 76.7% success rate. |
| 28 | * **[2026-01-13]** 🥇 **MAI-UI Sweeps ScreenSpot-Pro**: MAI-UI (32B, 8B, 2B) now ranks #1 in all size categories on the [ScreenSpot-Pro leaderboard](https://gui-agent.github.io/grounding-leaderboard/). We achieved record scores of 67.9%, 65.7%, and 57.4% respectively—notably reaching these benchmarks **without any zoom-in tricks**. |
| 29 | * **[2026-01-04]** 🤝 We're Hiring! We're actively looking for Research Scientists, Engineers, and Interns to work on foundational GUI agents and their applications. Interested candidates please send your resume to: yue.w@alibaba-inc.com |
| 30 | * **[2025-12-29]** 🏆 **New Leaderboard Record**: MAI-UI achieves a 41.7% success rate on the [MobileWorld](https://tongyi-mai.github.io/MobileWorld/#leaderboard) benchmark, setting a new record for end-to-end model performance! |
| 31 | * **[2025-12-29]** 📄 **Technical Report & Website**: Our technical report is now available on [arXiv](https://arxiv.org/abs/2512.22047), and the official project [website](https://tongyi-mai.github.io/MAI-UI-blog/) is live. |
| 32 | * **[2025-12-29]** 🤗 **Model Release**: We are excited to release the weights for [MAI-UI-8B](https://huggingface.co/Tongyi-MAI/MAI-UI-8B) and [MAI-UI-2B](https://huggingface.co/Tongyi-MAI/MAI-UI-2B) on Hugging Face. |
| 33 | |
| 34 | ## 📑 Table of Contents |
| 35 | |
| 36 | - [🎥 Demo](#-demo) |
| 37 | - [🚀 Quick Start](#-installation--quick-start) |
| 38 | - [📝 Citation](#-citation) |
| 39 | - [📧 Contact](#-contact) |
| 40 | - [📄 License](#-license) |
| 41 | |
| 42 | |
| 43 | <!-- |
| 44 | ## 🏆 Results |
| 45 | |
| 46 | MAI-UI establishes new state-of-the-art across GUI grounding and mobile navigation. |
| 47 | |
| 48 | - On grounding benchmarks, it reaches 73.5% on ScreenSpot-Pro, 91.3% on MMBench GUI L2, 70.9% on OSWorld-G, and 49.2% on UI-Vision, surpassing Gemini-3-Pro and Seed1.8 on ScreenSpot-Pro. |
| 49 | |
| 50 | <table align="center"> |
| 51 | <tr> |
| 52 | <td align="center |