GUI Grounding on OSWorld-G
72.7Average ScoreQwen3VL-8B-Instruct
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Qwen3VL-8B-InstructAccess=Open-Source, MVP Framework=true2025.12 | 72.7 | 82.4 | 76.7 | 79.5 | 61.7 | — | — | — | — | — | |
| Qwen3VL-32B-InstructAccess=Open-Source, MVP Framework=true2025.12 | 72 | 81.6 | 76.4 | 77.5 | 62.4 | — | — | — | — | — | |
| Qwen3VL-32B-InstructAccess=Open-Source, MVP Framework=false2025.12 | 71.7 | 81.6 | 75.7 | 77.1 | 62.4 | — | — | — | — | — | |
| UI-Venus-1.5-30B-A3Bdesign=end-to-end2026.02 | 70.6 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-1.5-30B-A3BModel Category=GUI-specific Models, Parameters=30B-A3B, Version=1.52026.06 | 70.6 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-Ground-72BAccess=Open-Source2025.12 | 70.4 | 82.1 | 71.2 | 70.7 | 64.4 | — | — | — | — | — | |
| Qwen3-VL-32B-Instruct + Re-Prefillzoom-in strategy=true, Re-Prefill=true2026.05 | 70.1 | 78.9 | 74.8 | 77.5 | 57 | — | — | — | — | — | |
| Step-GUI-8BParameters=8B2025.12 | 70 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-1.5-8Bdesign=end-to-end2026.02 | 69.7 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-1.5-8BModel Category=GUI-specific Models, Parameters=8B, Version=1.52026.06 | 69.7 | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-32B-Instructzoom-in strategy=true2026.05 | 69 | 76.6 | 73.9 | 77.9 | 54.4 | — | — | — | — | — | |
| Qwen3VL-8B-InstructAccess=Open-Source, MVP Framework=false2025.12 | 68.8 | 79.7 | 71.8 | 74.3 | 59.1 | — | — | — | — | — | |
| GUI-Owl-1.5-8B-Instruct + Re-Prefillzoom-in strategy=true, Re-Prefill=true2026.05 | 68.8 | 77.4 | 73 | 74.3 | 59.7 | — | — | — | — | — | |
| GTA1-7BAccess=Open-Source, MVP Framework=true2025.12 | 68.7 | 80.8 | 71.6 | 71.1 | 61 | — | — | — | — | — | |
| MAI-UI-8B + Re-Prefillzoom-in strategy=true, Re-Prefill=true2026.05 | 67.7 | 76.6 | 71.8 | 74.7 | 55.7 | — | — | — | — | — | |
| GTA1-7BInitialised from=UI-TARS-1.5-7B2025.09 | 67.7 | 63.2 | 82.1 | 74.2 | 42.9 | — | 0 | — | — | — | |
| EvoCUA-OpenCUA-72BBase Model=OpenCUA-72B2026.01 | 67.65 | — | — | — | — | — | — | — | — | — | |
| MAI-UI-32Bdesign=end-to-end2026.02 | 67.6 | — | — | — | — | — | — | — | — | — | |
| Mimo-VL-7B + WildGUIModel Category=Effectiveness of WildGUI (Ours), Training=Continual Pre-training on WildGUI2026.05 | 67.6 | 80.8 | 68.3 | 71.1 | 61.4 | — | — | — | — | — | |
| MAI-UI-32BModel Category=GUI-specific Models, Parameters=32B2026.06 | 67.6 | 73.6 | 72.4 | 73.9 | 57.7 | — | — | — | — | — | |
| GTA1-7BAccess=Open-Source, MVP Framework=false2025.12 | 67.5 | 80.4 | 69.4 | 69.1 | 60.4 | — | — | — | — | — | |
| GUI-Owl-1.5-8B-Instructzoom-in strategy=true2026.05 | 67.2 | 75.5 | 71.8 | 73.5 | 56.4 | — | — | — | — | — | |
| OpenCUA-72BParameters=72B2026.01 | 66.95 | — | — | — | — | — | — | — | — | — | |
| Step-GUI-4BParameters=4B2025.12 | 66.9 | — | — | — | — | — | — | — | — | — | |
| UI-TARS-1.5-7BAccess=Open-Source, MVP Framework=true2025.12 | 66.8 | 78.5 | 69.1 | 70 | 59.7 | — | — | — | — | — | |
| GUI-Owl-1.5-32B-InstructModel Category=GUI-specific Models, Parameters=32B, Version=1.5, Type=Instruct2026.06 | 66.8 | 68.2 | 66.8 | 65.5 | 62.1 | — | 70.4 | — | — | — | |
| GTA1-72BAccess=Open-Source2025.12 | 66.7 | 57.9 | 76.9 | 77.3 | 66.7 | — | — | — | — | — | |
| GUI-Owl-1.5-8B-Instruct2026.05 | 65.8 | 67.8 | 68.5 | 68.5 | 65.5 | — | 42.6 | — | — | — | |
| GUI-Owl-1.5-8B-InstructModel Category=GUI-specific Models, Parameters=8B, Version=1.5, Type=Instruct2026.06 | 65.8 | 67.8 | 68.5 | 68.5 | 65.5 | — | 42.6 | — | — | — | |
| Qwen3-VL-8B-Instruct + Re-Prefillzoom-in strategy=true, Re-Prefill=true2026.05 | 65.7 | 75.5 | 69.7 | 71.2 | 54.4 | — | — | — | — | — | |
| GUI-CursorInitialised from=UI-TARS-1.5-7B2025.09 | 65.6 | 77 | 66.7 | 70.4 | 67.4 | — | 0 | — | — | — | |
| MAI-UI-8Bzoom-in strategy=true2026.05 | 65.5 | 75.5 | 70.3 | 70 | 53.7 | — | — | — | — | — | |
| GTA1-32Bdesign=end-to-end2026.02 | 65.2 | — | — | — | — | — | — | — | — | — | |
| Holo2-30B-A3Breproduction=true, design=end-to-end2026.02 | 65.2 | — | — | — | — | — | — | — | — | — | |
| GTA1-32B2026.05 | 65.2 | 63.2 | 78.4 | 73.3 | 65.2 | — | — | — | — | — | |
| MAI-UI-8B2026.05 | 64.2 | 72.8 | 67.6 | 71.1 | 56.4 | — | — | — | — | — | |
| Qwen3-VL-32B-ThinkingParameters=32B, Thinking Mode=true2026.01 | 64 | — | — | — | — | — | — | — | — | — | |
| EvoCUA-32BParameters=32B2026.01 | 63.86 | — | — | — | — | — | — | — | — | — | |
| GUI-Owl-1.5-4B-InstructModel Category=GUI-specific Models, Parameters=4B, Version=1.5, Type=Instruct2026.06 | 63.7 | 66.1 | 66.1 | 64.8 | 64.8 | — | 35.2 | — | — | — | |
| Holo2-8Breproduction=true, design=end-to-end2026.02 | 63.5 | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8B-Instructzoom-in strategy=true2026.05 | 63 | 72.8 | 66.7 | 67.2 | 53.7 | — | — | — | — | — | |
| Seed1.5-VLAccess=Closed-source2025.12 | 62.9 | 73.9 | 66.7 | 69.6 | 47 | — | 18.5 | — | — | — | |
| Seed1.5-VLModel Category=Proprietary Models2026.05 | 62.9 | 73.9 | 66.7 | 69.6 | 47 | — | — | — | — | — | |
| GUI-SpotlightAccess=Open-Source2025.12 | 62.7 | 68.2 | 60.6 | 63.2 | 45.6 | — | — | — | — | — | |
| GUI-SpotlightInitialised from=UI-TARS-1.5-7B2025.09 | 62.7 | 68.2 | 60.6 | 63.2 | 45.6 | — | 0 | — | — | — | |
| UI-Venus-1.0-72Bdesign=end-to-end2026.02 | 62.2 | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-72BAccess=Open-Source2025.12 | 62.2 | 52.6 | 74.6 | 74.7 | 55.3 | — | — | — | — | — | |
| UI-Venus-Ground-72BModel Category=GUI-specific Models, Parameters=72B2026.06 | 62.2 | — | — | — | — | — | — | — | — | — | |
| UI-TARS-1.5-7BAccess=Open-Source, MVP Framework=false2025.12 | 61.9 | 67.3 | 64.5 | 65.2 | 42.9 | — | — | — | — | — | |
| UI-TARS-1.5-7B2025.09 | 61.9 | 67.3 | 64.5 | 65.2 | 42.9 | — | 0 | — | — | — | |
| Qwen3-VL-30B-A3Breproduction=true, design=end-to-end2026.02 | 61.2 | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-30B-A3BModel Category=General Models, Parameters=30B-A3B2026.06 | 61.2 | — | — | — | — | — | — | — | — | — | |
| Qwen3-30B-A3B-InstructParameters=30B, Arch=A3B, Type=Instruct2025.12 | 61 | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-32BModel Category=Open-Source Models, Evaluated By Authors=true2026.05 | 60.6 | 72.8 | 63.3 | 66.4 | 51.7 | — | — | — | — | — | |
| Step-GUI-4Breproduction=true, design=end-to-end2026.02 | 60.5 | — | — | — | — | — | — | — | — | — | |
| Step-GUI-4BModel Category=GUI-specific Models, Parameters=4B2026.06 | 60.5 | — | — | — | — | — | — | — | — | — | |
| GTA1-7Bdesign=end-to-end2026.02 | 60.1 | — | — | — | — | — | — | — | — | — | |
| MAI-UI-8Bdesign=end-to-end2026.02 | 60.1 | — | — | — | — | — | — | — | — | — | |
| MAI-UI-8BModel Category=GUI-specific Models, Parameters=8B2026.06 | 60.1 | 72 | 63.3 | 66 | 51 | — | — | — | — | — | |
| OpenCUA-32Bdesign=end-to-end2026.02 | 59.6 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-1.5-2Bdesign=end-to-end2026.02 | 59.4 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-7BModel Category=Open-Source Models2026.05 | 58.8 | 74.6 | 60.5 | 61.5 | 45.5 | — | — | — | — | — | |
| Xiaomi-GUI-0-30B-A3BModel Category=Ours, Parameters=30B-A3B2026.06 | 58.7 | 47.4 | 65.7 | 73.8 | 51.5 | — | 0 | — | — | — | |
| Trifuse+GUI-AIMAModel Size=3B2026.02 | 58.4 | 65.6 | 63.7 | 69.4 | 36.3 | 0.1 | — | — | — | — | |
| GUI-AIMATraining Data Size=85K, Model Size=3B2026.02 | 58.3 | 64.8 | 65.5 | 68.8 | 36.8 | — | — | — | — | — | |
| Qwen3-VL-4B-InstructParameters=4B, Type=Instruct2025.12 | 58.2 | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8B-InstructParameters=8B, Type=Instruct2025.12 | 58.2 | — | — | — | — | — | — | — | — | — | |
| GUI-Owl-32Bdesign=end-to-end2026.02 | 58 | — | — | — | — | — | — | — | — | — | |
| GUI-Owl-32BParameters=32B2025.12 | 58 | — | — | — | — | — | — | — | — | — | |
| GUI-CursorInitialised from=Qwen2.5-VL-7B2025.09 | 58 | 70.1 | 58.5 | 62.1 | 52.3 | — | 0 | — | — | — | |
| GUI-Owl-32BModel Category=GUI-specific Models, Parameters=32B2026.06 | 58 | 67 | 64.5 | 67.2 | 45.6 | — | — | — | — | — | |
| Trifuse+GUI-ActorModel Size=7B2026.02 | 57.9 | 66.4 | 62.3 | 67.1 | 38.9 | 1.3 | — | — | — | — | |
| GUI-Owl-1.5-32B-ThinkingModel Category=GUI-specific Models, Parameters=32B, Version=1.5, Type=Thinking2026.06 | 57.6 | 63.6 | 63.9 | 67 | 55.2 | — | 5.6 | — | — | — | |
| Qwen3-VL-8Breproduction=true, design=end-to-end2026.02 | 57.5 | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8BModel Category=General Models, Parameters=8B2026.06 | 57.5 | — | — | — | — | — | — | — | — | — | |
| UI-TARS-72BAccess=Open-Source2025.12 | 57.1 | 69.4 | 60.6 | 62.9 | 45.6 | — | — | — | — | — | |
| UI-TARS-72B2026.05 | 57.1 | 69.4 | 60.6 | 62.9 | 45.6 | — | — | — | — | — | |
| UI-TARS-72BModel Category=Open-Source Models2026.05 | 57.1 | 69.4 | 60.6 | 62.9 | 45.6 | — | — | — | — | — | |
| UI-TARS-72BModel Category=GUI-specific Models, Parameters=72B2026.06 | 57.1 | 69.4 | 60.6 | 62.9 | 45.6 | — | 0 | — | — | — | |
| Qwen3-VL-8B-ThinkingParameters=8B, Thinking Mode=true2026.01 | 56.7 | — | — | — | — | — | — | — | — | — | |
| GUI-ActorTraining Data Size=1M, Model Size=7B2026.02 | 56.6 | 65.9 | 62.7 | 66.4 | 38.2 | — | — | — | — | — | |
| Trifuse+GUI-ActorModel Size=3B2026.02 | 56.5 | 64.9 | 61.4 | 66.7 | 34.4 | 1.9 | — | — | — | — | |
| GUI-Owl-7Bdesign=end-to-end2026.02 | 55.9 | — | — | — | — | — | — | — | — | — | |
| GUI-Owl-7BParameters=7B2025.12 | 55.9 | — | — | — | — | — | — | — | — | — | |
| GUI-Owl-7BModel Category=Open-Source Models2026.05 | 55.9 | 64.8 | 63.6 | 61.3 | 41 | — | — | — | — | — | |
| GUI-Owl-7BModel Category=GUI-specific Models, Parameters=7B2026.06 | 55.9 | 64.8 | 63.6 | 61.3 | 41 | — | — | — | — | — | |
| OpenCUA-7Bdesign=end-to-end2026.02 | 55.3 | — | — | — | — | — | — | — | — | — | |
| OpenCUA-7BModel Category=Open-Source Models2026.05 | 55.3 | — | — | — | — | — | — | — | — | — | |
| GTA1-7BModel Category=Open-Source Models2026.05 | 55.1 | 42.1 | 65.7 | 62.7 | 56.1 | — | — | — | — | — | |
| EvoCUA-8BParameters=8B2026.01 | 55.08 | — | — | — | — | — | — | — | — | — | |
| GUI-Owl-1.5-8B-ThinkingModel Category=GUI-specific Models, Parameters=8B, Version=1.5, Type=Thinking2026.06 | 55 | 56.9 | 60 | 60.7 | 53.8 | — | 22.2 | — | — | — | |
| Qwen3-VL-8BModel Category=Open-Source Models, Evaluated By Authors=true2026.05 | 54.8 | 69 | 55.5 | 59.7 | 47.7 | — | — | — | — | — | |
| Mimo-VL-7BModel Category=Effectiveness of WildGUI (Ours)2026.05 | 54.7 | 65 | 59.2 | 59 | 40.2 | — | — | — | — | — | |
| GUI-ActorTraining Data Size=1M, Model Size=3B2026.02 | 54.6 | 64.4 | 60.6 | 64.8 | 33.6 | — | — | — | — | — | |
| UI-Venus-1.0-7Bdesign=end-to-end2026.02 | 54.6 | — | — | — | — | — | — | — | — | — | |
| UI-Venus-Ground-7BModel Category=GUI-specific Models, Parameters=7B2026.06 | 54.6 | — | — | — | — | — | — | — | — | — | |
| FocusUI-7BData Size=1M2026.03 | 54.4 | 63.6 | 61.2 | 63.6 | 34.9 | — | 0 | — | — | — | |
| ZoomUI-7BData Size=02026.03 | 54.2 | 57.9 | 55.2 | 55.7 | 43 | — | 1.1 | — | — | — | |
| JEDITraining Data Size=1.39M, Model Size=7B2026.02 | 54.1 | 65.9 | 55.5 | 57.7 | 46.9 | — | — | — | — | — | |
| Jedi-7BAccess=Open-Source2025.12 | 54.1 | 65.9 | 55.5 | 57.7 | 46.9 | — | 7.4 | — | — | — |