List your company

GUI Grounding on OSWorld G-Refine v1.0 (test)

76.1Overall Success RateHolo2
56.65261.70166.7571.799Dec 26, 2025Jan 23, 2026Feb 20, 2026Mar 20, 2026Apr 17, 2026May 15, 2026Jun 12, 2026

Evaluation results

MethodLinks
Holo2Model Scale=≥30B2026.0676.1—————
MAI-UI-32BZoom-In=true2025.127579.379.784.264.4—
MAI-UI-32BZoom-In=false2025.1273.979.779.48161.7—
MAI-UI-8BZoom-In=true2025.1272.979.378.884.259.7—
VISTAModel Scale=≥30B, Inference Method=Multi-view (+ MVP)2026.0672.5—————
GTA1-32B2025.1272.263.283.684.470.50
GTA1Model Scale=≥30B2026.0672.2—————
VISTAModel Scale=≥30B, Inference Method=Single-view2026.0672—————
VISTAModel Scale=≈8B, Inference Method=Multi-view (+ MVP)2026.0670.9—————
VISTAModel Scale=≈8B, Inference Method=Single-view2026.0670.8—————
OpenCUA-32B2025.1270.263.279.984.962.17.4
OpenCUAModel Scale=≥30B2026.0670.2—————
Holo2Model Scale=≈8B2026.0670.1—————
Step-GUIModel Scale=≈8B2026.0670—————
VISTAModel Scale=≈4B, Inference Method=Multi-view (+ MVP)2026.0669.6—————
UI-VenusModel Scale=72B2026.0669.5—————
VISTAModel Scale=≈4B, Inference Method=Single-view2026.0669.4—————
Qwen3-VLModel Scale=≥30B2026.0669.3—————
Qwen3-VL-32BEvaluated by authors=true2025.126977.473.676.357.7—
MAI-UI-8BZoom-In=false2025.1268.677.47378.355.7—
MAI-UIModel Scale=≈8B2026.0668.6—————
Qwen3-VLModel Scale=≈4B2026.0667.9—————
GTA1-7B2025.1267.763.282.174.270.50
GTA1Model Scale=≈8B2026.0667.7—————
Step-GUIModel Scale=≈4B2026.0666.9—————
MAI-UI-2BZoom-In=true2025.1266.371.371.878.351—
Qwen3-VL-8BEvaluated by authors=true2025.1264.473.968.273.154.4—
Qwen3-VLModel Scale=≈8B2026.0664.4—————
UI-TARS-1.5-7B2025.1264.252.675.472.466.70
UI-TARS-1.5Model Scale=≈8B2026.0664.2—————
Jedi-7B2025.1263.8—————
MAI-UI-2BZoom-In=false2025.1263.570.969.172.747.7—
UI-VenusModel Scale=≈8B2026.0661.7—————
Jedi-3B2025.1261—————
Qwen2.5-VL-32B2025.1259.657.970.273.849.20
Operator2025.1257.8—————
Qwen3-VL-2BEvaluated by authors=true2025.1257.469.360.969.245—