List your company

Mobile Task Automation on AndroidWorld (test)

1Average Success RateMinitap
0.31880.495650.67250.84935Apr 27, 2025Jun 21, 2025Aug 15, 2025Oct 10, 2025Dec 4, 2025Jan 28, 2026Mar 25, 2026

Evaluation results

MethodLinks
MinitapType=Agent2026.021———————————————————
AGI-0Type=Agent, Closed-source=true2026.020.974———————————————————
askui AndroidVisionAgentType=Agent2026.020.948———————————————————
DroidRunType=Agent2026.020.914———————————————————
Agent-ViscoType=Agent, Closed-source=true2026.020.888———————————————————
DroidRun(v0.3.9)+AppCardsModel=GPT-5, Screen Representation=Screenshot + A11y tree2026.010.888———————————————————
gbox.aiType=Agent2026.020.862———————————————————
mobile-useModel=Llama 4-scout, Gemini 2.5 pro, GPT-5 nano, Screen Representation=Screenshot + A11y tree2026.010.845———————————————————
UI-Voyager#Params=4B2026.030.81———————————————————
AutoGLM-MobileType=Model, Parameters=9B2026.020.802———————————————————
AutoGLM-MobileModel=AutoGLM-Mobile, Screen Representation=Screenshot + A11y tree2026.010.802———————————————————
HumanType=Human baseline2026.020.8———————————————————
Human#Params=-2026.030.8———————————————————
LX-GUIAgentModel=LX-GUIAgent, Screen Representation=Screenshot + A11y tree2026.010.793———————————————————
DroidRun(v0.3.9)Model=GPT-5, Gemini 2.5 Pro, Screen Representation=Screenshot + A11y tree2026.010.784———————————————————
UI-Venus-1.5-30B-A3B#Params=30B2026.030.776———————————————————
MAI-UIType=Model, Parameters=235B2026.020.767———————————————————
FinalRunType=Training free, Base Model=GPT-5, Input=Screenshot + A11y2026.020.767———————————————————
MAI-UI-235B-A22B#Params=235B2026.030.767———————————————————
K2-AgentType=Ours, Base Model=Qwen2.5-VL (72B+7B), Input=Screenshot2026.020.761———————————————————
AutoGLM-MobileType=Learning based, Base Model=AutoGLM-Mobile, Input=Screenshot + A11y2026.020.758———————————————————
MobileUse-v2Model=Hammer-UI-32B, Screen Representation=Screenshot + A11y tree2026.010.75———————————————————
Mobile-Agent-v3Model=GUI-Owl-32B, Screen Representation=Screenshot2026.010.733———————————————————
Mobile-Agent-v3Type=Learning based, Base Model=Qwen-VL based, Input=Screenshot2026.020.733———————————————————
UI-TARS-2Type=Learning based, Base Model=Seed-thinking-1.6, Input=Screenshot2026.020.733———————————————————
MAI-UI-32B#Params=32B2026.030.733———————————————————
UI-Tars-2#Params=230B2026.030.733———————————————————
DroidRun(v0.3.9)+AppCardsModel=Gemini-2.5-Pro, Screen Representation=Screenshot + A11y tree2026.010.716———————————————————
GUI-Owl-1.5-8B-Thinking#Params=8B2026.030.716———————————————————
MAI-UIType=Model, Parameters=8B2026.020.707———————————————————
MAI-UI-8B#Params=8B2026.030.707———————————————————
Seed1.8#Params=-2026.030.707———————————————————
Gemini 2.5 ComputerType=Model, Closed-source=true2026.020.697———————————————————
Gemini-2.5-Pro#Params=-2026.030.697———————————————————
DroidRun(v0.3.3)+AppCardsModel=Gemini-2.5-Pro, Screen Representation=Screenshot + A11y tree2026.010.69———————————————————
Step-GUI-8B#Params=8B, Reference=Yan et al., 2025a2026.030.677———————————————————
Step-GUI-8B#Params=8B, Reference=Yan et al., 2025b2026.030.677———————————————————
JT-GUIAgent-V2Model=JT-GUIAgent-V2, Screen Representation=Screenshot2026.010.672———————————————————
GUI-Owl-7BType=Model, Parameters=7B2026.020.664———————————————————
DMS-PA-LiteUniv.=true, Weights=true, Train.=true, Mem.=true, Model=Qwen2.5-VL-72B-Instruct2026.010.664———————————————————
GUI-Owl-7B#Params=7B2026.030.664———————————————————
UI-VenusType=Model, Parameters=72B2026.020.659———————————————————
UI-VenusUniv.=true, Weights=true, Train.=false, Mem.=false, Model=Qwen2-VL-72B-Instruct2026.010.659———————————————————
UI-VenusType=Learning based, Base Model=Qwen2.5-VL-72B, Input=Screenshot2026.020.659———————————————————
UI-Venus-72B#Params=72B2026.030.659———————————————————
UI-TARS 1.52025.050.642———————————————————
UI-Tars-1.5#Params=-2026.030.642———————————————————
Step-GUI-4B#Params=4B2026.030.639———————————————————
Qwen3-VL-235B-A22B#Params=235B2026.030.637———————————————————
DroidRun(v0.3.3)Model=Gemini-2.5-Pro, Screen Representation=Screenshot + A11y tree2026.010.63———————————————————
DroidRunType=Training free, Base Model=Gemini-2.5-Pro, Input=Screenshot + A11y2026.020.63———————————————————
MobileUseUniv.=false, Weights=true, Train.=true, Mem.=true, Model=Qwen2.5-VL-72B-Instruct2026.010.629———————————————————
MobileUseType=Training free, Base Model=Qwen2.5-VL-72B, Input=Screenshot2026.020.629———————————————————
Seed 1.5-VL2025.050.621———————————————————
Seed1.5-VLType=Model, Parameters=20B2026.020.621———————————————————
Seed1.5-VLUniv.=true, Weights=true, Train.=true, Mem.=false, Model=Seed1.5-VL2026.010.621———————————————————
Seed1.5-VLType=Learning based, Base Model=Seed1.5-VL-72B, Input=Screenshot + A11y2026.020.621———————————————————
Seed1.5-VL#Params=-2026.030.621———————————————————
DroidRun(v0.3.9)+AppCardsModel=Grok-4-Fast, Screen Representation=Screenshot + A11y tree2026.010.603———————————————————
JT-GUIAgent-V1Model=JT-GUIAgent-V1, Screen Representation=Screenshot2026.010.6———————————————————
V-DroidUniv.=false, Weights=true, Train.=false, Mem.=true, Model=V-Droid2026.010.595———————————————————
V-DroidModel=V-Droid (Llama-8B), Screen Representation=A11y tree2026.010.595———————————————————
Qwen3-VL-32B#Params=32B2026.030.573———————————————————
DMS-PA-LiteUniv.=true, Weights=false, Train.=true, Mem.=true, Model=Seed1.6-VL2026.010.56———————————————————
Agent-S2Univ.=true, Weights=false, Train.=true, Mem.=true, Model=Claude-3.7-Sonnet2026.010.543———————————————————
Agent S2Model=Agent S2, Screen Representation=Screenshot2026.010.543———————————————————
Agent S2Type=Training free, Base Model=Claude-3.5-Sonnet, Input=Screenshot2026.020.543———————————————————
EchoTrail-GUIUniv.=false, Weights=false, Train.=true, Mem.=true, Model=GPT-4o2026.010.517———————————————————
DMS-PA-LiteUniv.=true, Weights=true, Train.=true, Mem.=true, Model=GLM-4.5V2026.010.504———————————————————
MAI-UI-2B#Params=2B2026.030.491———————————————————
UI-Venus-7B#Params=7B2026.030.491———————————————————
Qwen3-VL-8B#Params=8B2026.030.476———————————————————
GUI-ExplorerModel=GPT-4o, Screen Representation=Screenshot + A11y tree2026.010.474———————————————————
ANDROIDGENBase Model=GPT-4o, Fine-tuned=false2025.040.46810.125100.6670.6670.3750.510.3570.1670.3330.4550.50.9330.5710.1670.5—
AndroidGenUniv.=false, Weights=false, Train.=true, Mem.=true, Model=GPT-4o2026.010.468———————————————————
AndroidGenModel=GPT-4o, Screen Representation=A11y tree2026.010.468———————————————————
AndroidGenType=Training free, Base Model=GPT-o, Input=A11y Tree2026.020.468———————————————————
UI-TARSType=Model, Parameters=72B2026.020.466———————————————————
EchoTrail-GUIUniv.=false, Weights=true, Train.=true, Mem.=true, Model=Qwen2.5-VL-72B-Instruct2026.010.466———————————————————
UI-TARSUniv.=true, Weights=true, Train.=false, Mem.=false, Model=Qwen2-VL-72B-Instruct2026.010.466———————————————————
UI-TARSModel=UI-TARS, Screen Representation=Screenshot2026.010.466———————————————————
UI-TARSType=Learning based, Base Model=Qwen-2-VL-72B, Input=Screenshot2026.020.466———————————————————
UI-Tars-SFT-72B#Params=72B2026.030.466———————————————————
Qwen3-VL-4B#Params=4B2026.030.453———————————————————
MobileGUI-RLType=Model, Parameters=32B2026.020.448———————————————————
Aria-UIType=Model2026.020.448———————————————————
Aria-UIUniv.=true, Weights=false, Train.=false, Mem.=false, Model=GPT-4o + Aria-UI2026.010.448———————————————————
Aria-UIModel=GPT-4o + Aria-UI, Screen Representation=Screenshot2026.010.448———————————————————
UGroundType=Model2026.020.44———————————————————
UGroundUniv.=false, Weights=false, Train.=true, Mem.=false, Model=GPT-4o2026.010.44———————————————————
DMS-PA-LiteUniv.=true, Weights=true, Train.=true, Mem.=true, Model=Qwen3-VL 30B-A3B-Instruct2026.010.44———————————————————
PA-LiteUniv.=true, Weights=false, Train.=true, Mem.=false, Model=Seed1.6-VL2026.010.394——————————————————1.3
PA-LiteUniv.=true, Weights=true, Train.=true, Mem.=false, Model=Qwen2.5-VL-72B-Instruct2026.010.392——————————————————1.5
AguvisUniv.=true, Weights=false, Train.=false, Mem.=false, Model=GPT-4o + Aguvis2026.010.371———————————————————
Qwen3-VL-2B#Params=2B2026.030.364———————————————————
ANDROIDGENBase Model=Llama-3-70B, Fine-tuned=true2025.040.35310.063100.6670.3330.3330.50.50.2860.16700.40.250.70.66700—
PA-LiteUniv.=true, Weights=true, Train.=true, Mem.=false, Model=GLM-4.5V2026.010.352——————————————————1.9
Qwen 2.5 VL 72BParameters=72B2025.050.35———————————————————
Qwen2.5-VLUniv.=true, Weights=true, Train.=true, Mem.=false, Model=Qwen2.5-VL-72B-Instruct2026.010.35———————————————————
GPT-4oUniv.=true, Weights=false, Train.=true, Mem.=false, Model=GPT-4o2026.010.345———————————————————