UI-TARS-72B-DPO
Multimodal
by ByteDance
+
+
+
+
About
UI-TARS-72B-DPO is a model from ByteDance's UI-TARS family built on the Qwen base and trained for GUI interaction tasks through direct preference optimization. It represents ByteDance's early work establishing the UI-TARS architecture as a specialized model family for multimodal GUI agent research — a line of models focused on making AI agents that interact with software interfaces through visual perception rather than programmatic API access.
+
+
+
+
Timeline
ReleasedJan 22, 2025
+
+
+
+
Specifications
Capabilities
Multimodal
+
+
+
+
License & Family
License
Apache 2.0
Performance Overview
Performance metrics and category breakdown
No benchmark data available for this model