GUI Agent Task on AndroidWorld
80Success RateHuman
Evaluation results
| Method | Links | |
|---|---|---|
| HumanInput=screen2025.05 | 80 | |
| AgentProgBackbone Model=Gemini-2.5-Pro, Element Locator=UI-TARS-1.5-API2025.12 | 78 | |
| LAMO-3BExecution Mode=Policy Executor Mode, Planner=GPT-52026.04 | 77.6 | |
| UI-KOBEAgent Type=Our, Size / Model=Qwen3.5-Plus2026.05 | 77.6 | |
| MAI-UI-235B-A22BAgent Type=Single Model, Size / Model=235B2026.05 | 76.7 | |
| ATMem-UIParams=8B2026.06 | 76.6 | |
| UI-Ins-7BPlanner=GPT-52026.04 | 74.1 | |
| Mobile-Agent-v3Backbone Model=GUI-Owl-32B2025.12 | 73.3 | |
| UI-TARS-2Params=-2026.03 | 73.3 | |
| Mobile-Agent-V32026.04 | 73.3 | |
| MAI-UI-32BAgent Type=Single Model, Size / Model=32B2026.05 | 73.3 | |
| Mobile-Agent-v3Agent Type=Agentic Framework, Size / Model=GUI-Owl-32B2026.05 | 73.3 | |
| UI-TARS-2Params=230B2026.06 | 73.3 | |
| UI-KOBEAgent Type=Our, Size / Model=Qwen3.5-9B2026.05 | 72.4 | |
| GUI-Owl-1.5-ThinkingParams=8B2026.06 | 71.6 | |
| MAI-UI-8BAgent Type=Single Model, Size / Model=8B2026.05 | 70.7 | |
| UI-KOBEAgent Type=Our, Size / Model=Qwen3.5-4B2026.05 | 70.7 | |
| Seed1.82026.06 | 70.7 | |
| MAI-UIParams=8B2026.06 | 70.7 | |
| Gemini-2.5-Pro2026.06 | 69.7 | |
| Agent S32025.12 | 68.1 | |
| Step-GUI-8BAgent Type=Single Model, Size / Model=8B2026.05 | 67.7 | |
| Step-GUIParams=8B2026.06 | 67.7 | |
| JT-GUIAgentV2Platform=Mobile2026.04 | 67.2 | |
| Qwen3.5-plusAgent Type=Single Model, Size / Model=397B2026.05 | 66.8 | |
| DMS-PA-LiteUnified Backbone=true, Open Weights=true, Training-Free=true, Evolutionary Memory=true, Model=Qwen2.5-VL-72B-Instruct2026.01 | 66.4 | |
| Mobile-Agent-v3Backbone Model=GUI-Owl-7B2025.12 | 66.4 | |
| GUI-Owl-7BAgent Type=Single Model, Size / Model=7B2026.05 | 66.4 | |
| UI-VenusUnified Backbone=true, Open Weights=true, Training-Free=false, Evolutionary Memory=false, Model=Qwen2.5-VL-72B-Instruct†2026.01 | 65.9 | |
| UI-Venus2025.12 | 65.9 | |
| UI-Venus-Navi-72B2026.04 | 65.9 | |
| UI-Venus-72BAgent Type=Single Model, Size / Model=72B2026.05 | 65.9 | |
| UI-TARS-SFTParams=72B2026.06 | 65.9 | |
| UI-VenusParams=72B2026.06 | 65.9 | |
| ATMem-UIParams=4B2026.06 | 65.5 | |
| Qwen3-VL-32B-Thinking w/ TraceR1Params=32B w/ 8B2026.03 | 64.8 | |
| UI-TARSModel Scale/Version=UI-TARS-1.5-API2025.12 | 64.2 | |
| UI-TARS-1.5Params=-2026.03 | 64.2 | |
| UI-TARS-1.52026.06 | 64.2 | |
| Qwen3-VL-235B-A22B-Instruct2026.04 | 63.7 | |
| Qwen3-VL-235B-A22BAgent Type=Single Model, Size / Model=235B2026.05 | 63.7 | |
| Qwen3-VL-235B-A22BParams=235B2026.06 | 63.7 | |
| MobileUseUnified Backbone=false, Open Weights=true, Training-Free=true, Evolutionary Memory=true, Model=Qwen2.5-VL-72B-Instruct2026.01 | 62.9 | |
| MobileUse2025.12 | 62.9 | |
| MobileUseAgent Type=Agentic Framework, Size / Model=Qwen2.5-VL-72B2026.05 | 62.9 | |
| AppAgentXType=GPT-4o, Input=SoM2026.05 | 62.5 | |
| Seed1.5-VLUnified Backbone=true, Open Weights=true, Training-Free=true, Evolutionary Memory=false, Model=Seed1.5-VL2026.01 | 62.1 | |
| Seed1.5-VLParams=-2026.03 | 62.1 | |
| Qwen3-VL-32B-ThinkingParams=32B2026.03 | 61.4 | |
| LAMO-3BExecution Mode=Policy Executor Mode, Planner=Gemini-2.5-pro2026.04 | 60.3 | |
| V-DroidUnified Backbone=false, Open Weights=true, Training-Free=false, Evolutionary Memory=true, Model=V-Droid2026.01 | 59.5 | |
| V-Droid2025.12 | 59.5 | |
| V-DroidAgent Type=Agentic Framework, Size / Model=V-Droid-8B2026.05 | 59.5 | |
| Qwen3.5-4BAgent Type=Single Model, Size / Model=4B2026.05 | 58.6 | |
| Qwen3.5-9BAgent Type=Single Model, Size / Model=9B2026.05 | 57.8 | |
| Qwen3-VL-32BAgent Type=Single Model, Size / Model=32B2026.05 | 57.3 | |
| DMS-PA-LiteUnified Backbone=true, Open Weights=false, Training-Free=true, Evolutionary Memory=true, Model=Seed1.6-VL2026.01 | 56 | |
| GUI-Owl-7BBackbone=GUI-Owl-7B2026.01 | 54.3 | |
| Agent-S2Unified Backbone=true, Open Weights=false, Training-Free=true, Evolutionary Memory=true, Model=Claude-3.7-Sonnet2026.01 | 54.3 | |
| Agent S2 w/ Claude 3.7 SonnetParams=7B w/ -2026.03 | 54.3 | |
| Agent-S22026.04 | 54.3 | |
| Agent-S2Agent Type=Agentic Framework, Size / Model=Claude-3.7-sonnet2026.05 | 54.3 | |
| EAMType=GPT-4o, Qwen2.5-3B-instruct-ft, Input=SoM2026.05 | 52.6 | |
| OpenAI CUA-o3Params=-2026.03 | 52.5 | |
| OpenAI CUA-o32026.04 | 52.5 | |
| EchoTrail-GUIUnified Backbone=false, Open Weights=false, Training-Free=true, Evolutionary Memory=true, Model=GPT-4o2026.01 | 51.7 | |
| MobileExplorerSetting=fully on-device2026.05 | 50.9 | |
| DMS-PA-LiteUnified Backbone=true, Open Weights=true, Training-Free=true, Evolutionary Memory=true, Model=GLM-4.5V2026.01 | 50.4 | |
| Qwen3-VL-8B-ThinkingParams=8B2026.03 | 50 | |
| MAI-UI-2BAgent Type=Single Model, Size / Model=2B2026.05 | 49.1 | |
| UI-Venus-7BAgent Type=Single Model, Size / Model=7B2026.05 | 49.1 | |
| Qwen3-VL-8BAgent Type=Single Model, Size / Model=8B2026.05 | 47.6 | |
| GUI-explorerInput=SOM, Base Model=GPT-4o2025.05 | 47.4 | |
| GUI-ExplorerType=GPT-4o, Input=SoM2026.05 | 47.4 | |
| GUI-ExplorerAgent Type=Agentic Framework, Size / Model=GPT-4o2026.05 | 47.4 | |
| AndroidGenUnified Backbone=false, Open Weights=false, Training-Free=true, Evolutionary Memory=true, Model=GPT-4o2026.01 | 46.8 | |
| GPT-5Backbone=GPT-52026.01 | 46.6 | |
| EchoTrail-GUIUnified Backbone=false, Open Weights=true, Training-Free=true, Evolutionary Memory=true, Model=Qwen2.5-VL-72B-Instruct2026.01 | 46.6 | |
| UI-TARSUnified Backbone=true, Open Weights=true, Training-Free=false, Evolutionary Memory=false, Model=Qwen2.5-VL-72B-Instruct†2026.01 | 46.6 | |
| UI-TARS-72B2026.04 | 46.6 | |
| UI-Tars-72BAgent Type=Single Model, Size / Model=72B2026.05 | 46.6 | |
| MobileGenBackbone=Qwen3-VL-8B-Inst., Training Strategy=MobileGen2026.01 | 45.7 | |
| Aria-UITHInput=Image, Planner=GPT-4o2024.12 | 44.8 | |
| Aria-UIInput=screen, Base Model=GPT-4o2025.05 | 44.8 | |
| Aria-UIUnified Backbone=true, Open Weights=false, Training-Free=false, Evolutionary Memory=false, Model=GPT-4o + Aria-UI2026.01 | 44.8 | |
| GPT-4o + Aria-UI2026.04 | 44.8 | |
| SeeAct-VPlanner=GPT-4o, Grounding=UGround-V1-7B2024.10 | 44 | |
| UGroundUnified Backbone=false, Open Weights=false, Training-Free=true, Evolutionary Memory=false, Model=GPT-4o2026.01 | 44 | |
| DMS-PA-LiteUnified Backbone=true, Open Weights=true, Training-Free=true, Evolutionary Memory=true, Model=Qwen3-VL-30B-A3B-Instruct2026.01 | 44 | |
| ADMIRE(w/Qwen2.5-VL-7B-Instruct)#Params=7B, Base Model=Qwen2.5-VL-7B-Instruct2026.02 | 44 | |
| MAGNETBackbone=Qwen2.5-VL-32B, Memory-augmentation=true2026.01 | 42.62 | |
| MobileGenBackbone=InternVL3-14B, Training Strategy=MobileGen2026.01 | 42.2 | |
| GLM-4.1V-9B-ThinkingModel Scale=9B, Feature=Thinking2025.12 | 41.7 | |
| GLM-4.1V-9B-Thinking#Params=9B2026.02 | 41.7 | |
| GLM-4.1V-9B-ThinkingBackbone=GLM-4.1V-9B-Thinking2026.05 | 41.7 | |
| GUI-RewalkBackbone=Qwen3-VL-8B-Inst., Training Strategy=GUI-Rewalk2026.01 | 41.4 | |
| PA-LiteUnified Backbone=true, Open Weights=false, Training-Free=true, Evolutionary Memory=false, Model=Seed1.6-VL2026.01 | 41 | |
| PA-LiteUnified Backbone=true, Open Weights=true, Training-Free=true, Evolutionary Memory=false, Model=Qwen2.5-VL-72B-Instruct2026.01 | 41 | |
| Claude-Sonnet-4#Params=-2026.02 | 41 | |
| Agent-SBackbone=Qwen2.5-VL-32B, Memory-augmentation=true2026.01 | 40.98 |