MAI-UI Technical Report: Real-World Centric Foundation GUI Agents

Papers with Code Trending Papers
gui-agents foundation-models human-computer-interaction reinforcement-learning mobile-navigation privacy benchmarks

Summary

The MAI-UI technical report presents a family of foundation GUI agents in multiple sizes, addressing real-world deployment challenges with a self-evolving data pipeline, device-cloud collaboration, and online RL, achieving state-of-the-art results on GUI grounding and mobile navigation benchmarks.

The development of GUI agents could revolutionize the next generation of human-computer interaction. Motivated by this vision, we present MAI-UI, a family of foundation GUI agents spanning the full spectrum of sizes, including 2B, 8B, 32B, and 235B-A22B variants. We identify four key challenges to realistic deployment: the lack of native agent-user interaction, the limits of UI-only operation, the absence of a practical deployment architecture, and brittleness in dynamic environments. MAI-UI addresses these issues with a unified methodology: a self-evolving data pipeline that expands the navigation data to include user interaction and MCP tool calls, a native device-cloud collaboration system routes execution by task state, and an online RL framework with advanced optimizations to scale parallel environments and context length. MAI-UI establishes new state-of-the-art across GUI grounding and mobile navigation. 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. On mobile GUI navigation, it sets a new SOTA of 76.7% on AndroidWorld, surpassing UI-Tars-2, Gemini-2.5-Pro and Seed1.8. On MobileWorld, MAI-UI obtains 41.7% success rate, significantly outperforming end-to-end GUI models and competitive with Gemini-3-Pro based agentic frameworks. Our online RL experiments show significant gains from scaling parallel environments from 32 to 512 (+5.2 points) and increasing environment step budget from 15 to 50 (+4.3 points). Finally, the native device-cloud collaboration system improves on-device performance by 33%, reduces cloud model calls by over 40%, and preserves user privacy.
Original Article
View Cached Full Text

Cached at: 08/20/26, 09:54 PM

Paper page - MAI-UI Technical Report: Real-World Centric Foundation GUI Agents

Source: https://huggingface.co/papers/2512.22047

Abstract

ThedevelopmentofGUIagentscouldrevolutionizethenextgenerationofhuman-computerinteraction.Motivatedbythisvision,wepresentMAI-UI,afamilyoffoundationGUIagentsspanningthefullspectrumofsizes,including2B,8B,32B,and235B-A22Bvariants.Weidentifyfourkeychallengestorealisticdeployment:thelackofnativeagent-userinteraction,thelimitsofUI-onlyoperation,theabsenceofapracticaldeploymentarchitecture,andbrittlenessindynamicenvironments.MAI-UIaddressestheseissueswithaunifiedmethodology:aself-evolvingdatapipelinethatexpandsthenavigationdatatoincludeuserinteractionandMCPtoolcalls,anativedevice-cloudcollaborationsystemroutesexecutionbytaskstate,andanonlineRLframeworkwithadvancedoptimizationstoscaleparallelenvironmentsandcontextlength.MAI-UIestablishesnewstate-of-the-artacrossGUIgroundingandmobilenavigation.Ongroundingbenchmarks,itreaches73.5%onScreenSpot-Pro,91.3%onMMBenchGUIL2,70.9%onOSWorld-G,and49.2%onUI-Vision,surpassingGemini-3-ProandSeed1.8onScreenSpot-Pro.OnmobileGUInavigation,itsetsanewSOTAof76.7%onAndroidWorld,surpassingUI-Tars-2,Gemini-2.5-ProandSeed1.8.OnMobileWorld,MAI-UIobtains41.7%successrate,significantlyoutperformingend-to-endGUImodelsandcompetitivewithGemini-3-Probasedagenticframeworks.OuronlineRLexperimentsshowsignificantgainsfromscalingparallelenvironmentsfrom32to512(+5.2points)andincreasingenvironmentstepbudgetfrom15to50(+4.3points).Finally,thenativedevice-cloudcollaborationsystemimproveson-deviceperformanceby33%,reducescloudmodelcallsbyover40%,andpreservesuserprivacy.

View arXiv pageView PDFGitHub2.02kautoAdd to collection

Get this paper in your agent:

hf papers read 2512\.22047

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper6

#### Tongyi-MAI/MAI-UI-8B Image-Text-to-Text• 9B• UpdatedJan 9 • 2.89k • 199 #### mlx-community/MAI-UI-8B-bf16 Image-Text-to-Text• 9B• UpdatedMar 20 • 37 #### mlx-community/MAI-UI-8B-6bit Image-Text-to-Text• 2B• UpdatedMar 20 • 22 #### mlx-community/MAI-UI-8B-4bit Image-Text-to-Text• 2B• UpdatedMar 20 • 32 Browse 6 models citing this paper## Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2512.22047 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2512.22047 in a Space README.md to link it from this page.

Collections including this paper6

Browse 6 collections that include this paper

Similar Articles

Xiaomi-GUI-0 Technical Report

Hugging Face Daily Papers

This technical report presents Xiaomi-GUI-0, a native multimodal GUI agent trained and evaluated in real-device environments using a hybrid infrastructure and a progressive training pipeline, achieving high success rates and improved stability on real-world mobile tasks.

UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning

Papers with Code Trending

UI-TARS-2 is a native GUI-centered agent model that addresses data scalability, multi-turn RL, and environment stability challenges, achieving state-of-the-art results on GUI benchmarks (88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena,73.3 on AndroidWorld) and outperforming Claude and OpenAI agents.

Macaron-A2UI: A Model for Generative UI in Personal Agents

Hugging Face Daily Papers

Presents Macaron-A2UI, a model for generative UI in personal agents that synthesizes dynamic interfaces with lightweight executable actions, moving beyond text-only chat. The paper introduces a large-scale corpus, the A2UI-Bench benchmark, and trains models up to 754B parameters using LoRA fine-tuning and reinforcement learning, achieving strong results.