GUI Grounding on MMBench-GUI-L2
91.3AccuracyMAI-UI-32B
Evaluation Results
| Method | Links | |
|---|---|---|
| MAI-UI-32BParameters=32B2026.06 | 91.3 | |
| MAI-UI-8BParameters=8B2026.06 | 88.8 | |
| UI-Venus-1.5-30B-A3BVersion=1.5, Parameters=30B-A3B2026.06 | 88.6 | |
| UI-Venus-1.5-8BVersion=1.5, Parameters=8B2026.06 | 88.1 | |
| GUI-Owl-1.5-32B-InstructVersion=1.5, Parameters=32B, Variant=Instruct2026.06 | 86.8 | |
| GUI-SDTime/epoch=4.22026.05 | 86.7 | |
| UI-Venus-Ground-72BParameters=72B2026.06 | 86.3 | |
| Step-GUI-8BParameters=8B2025.12 | 85.6 | |
| Step-GUI-8BParameters=8B2026.06 | 85.6 | |
| GUI-Owl-1.5-32B-ThinkingVersion=1.5, Parameters=32B, Variant=Thinking2026.06 | 84.5 | |
| SeedVL-1.5Version=1.52025.12 | 84.4 | |
| Qwen3-VL-Instruct + GRPO-BinaryTime/epoch=16.92026.05 | 84.3 | |
| Step-GUI-4BParameters=4B2025.12 | 84 | |
| Step-GUI-4BParameters=4B2026.06 | 84 | |
| Qwen3-VL-Instruct + GRPO-GaussianTime/epoch=16.82026.05 | 83.7 | |
| Qwen3-VL-30B-A3BParameters=30B-A3B2026.06 | 83.7 | |
| Qwen3-VL-Instruct + GRPO-DistanceTime/epoch=16.72026.05 | 83.3 | |
| GUI-Owl-1.5-4B-InstructVersion=1.5, Parameters=4B, Variant=Instruct2026.06 | 83.2 | |
| UI-Ins-7B2026.05 | 83.1 | |
| GUI-Owl-32BParameters=32B2025.12 | 83 | |
| Qwen3-VL-Instruct2026.05 | 83 | |
| GUI-Owl-32BParameters=32B2026.06 | 83 | |
| Xiaomi-GUI-0-30B-A3BParameters=30B-A3B2026.06 | 82.7 | |
| GUI-Owl-1.5-8B-InstructVersion=1.5, Parameters=8B, Variant=Instruct2026.06 | 82.5 | |
| Qwen3-VL-8BParameters=8B2026.06 | 81.4 | |
| GUI-Owl-7BParameters=7B2025.12 | 80.5 | |
| GUI-Owl-7BParameters=7B2026.06 | 80.5 | |
| Qwen3-VL-8BModel Category=General Open-source Models2026.04 | 80.1 | |
| GUI-Owl-1.5-8B-ThinkingVersion=1.5, Parameters=8B, Variant=Thinking2026.06 | 80.1 | |
| UI-Venus-Ground-7BParameters=7B2026.06 | 79.9 | |
| Qwen3-30B-A3B-InstructParameters=30B, Arch=A3B, Type=Instruct2025.12 | 79.7 | |
| HyperClick-7B2026.05 | 79.6 | |
| Ours (Qwen3-VL-8B)Model Category=Ours2026.04 | 78.4 | |
| Propose-then-Critic-8B2026.05 | 78.4 | |
| Qwen3-VL-4BModel Category=General Open-source Models2026.04 | 77.4 | |
| UI-TARS-72BParameters=72B2026.06 | 74.3 | |
| UI-TARS-72B-DPOParameters=72B, Alignment=DPO2025.12 | 74.2 | |
| ZoomUI-7B2026.05 | 72.8 | |
| Ours (UI-Tars1.5)Model Category=Ours2026.04 | 68.9 | |
| UGround-V1-7BModel Category=GUI-specific Models2026.04 | 65.6 | |
| UI-TARS-1.5-7BVersion=1.5, Parameters=7B2025.12 | 64.3 | |
| UI-TARS-1.5-7BModel Category=GUI-specific Models2026.04 | 64.3 | |
| UI-TARS-1.5-7BVersion=1.5, Parameters=7B2026.06 | 64.3 | |
| GRPO w. Two.Annotation Noise Level=0%2025.11 | 60.6 | |
| GRPO w. Max.Annotation Noise Level=0%2025.11 | 58.2 | |
| GRPO w. Min.Annotation Noise Level=0%2025.11 | 58 | |
| GRPO w. Two.Annotation Noise Level=50%2025.11 | 57.4 | |
| GRPOAnnotation Noise Level=0%2025.11 | 57 | |
| GRPO w. Min.Annotation Noise Level=50%2025.11 | 55.6 | |
| GRPO w. Max.Annotation Noise Level=50%2025.11 | 54.6 | |
| GRPOAnnotation Noise Level=50%2025.11 | 53.8 | |
| GRPO w. Min.Annotation Noise Level=100%2025.11 | 51 | |
| GRPO w. Two.Annotation Noise Level=100%2025.11 | 49 | |
| GRPO w. Max.Annotation Noise Level=100%2025.11 | 48.8 | |
| Qwen3-VL-2BModel Category=General Open-source Models2026.04 | 48.1 | |
| GRPOAnnotation Noise Level=100%2025.11 | 46.4 | |
| Base ModelAnnotation Noise Level=Base2025.11 | 45 | |
| Qwen2.5-VL-72BVersion=2.5, Parameters=72B2025.12 | 41.8 | |
| OS-Atlas-Base-7BModel Category=GUI-specific Models2026.04 | 41.4 | |
| Qwen2.5-VL-7BModel Category=General Open-source Models2026.04 | 33.8 | |
| Qwen2.5-VL-3BModel Category=General Open-source Models2026.04 | 21.2 | |
| Claude-3.7 SonnetModel Category=Closed-Source Models2026.04 | 4.66 | |
| GPT-4oModel Category=Closed-Source Models2026.04 | 2.87 |