UI-TARS-2 advances GUI agent with multi-turn reinforcement learning

The UI-TARS-2 technical report introduces a native GUI-centered agent model that addresses challenges in data scalability, multi-turn reinforcement learning, and environment stability through a systematic training methodology. This approach includes a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment integrating file systems and terminals, and a unified sandbox platform.

• On GUI benchmarks, UI-TARS-2 achieves 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld, outperforming strong baselines like Claude and OpenAI agents.
• In game environments, it attains a mean normalized score of 59.8 across a 15-game suite, reaching roughly 60% of human-level performance and remaining competitive with frontier proprietary models such as OpenAI o3 on LMGame-Bench.
• The model demonstrates robustness by generalizing to long-horizon information-seeking tasks and software engineering benchmarks.

These results underscore UI-TARS-2's potential to advance the state of GUI agents and exhibit strong generalization to real-world interactive scenarios.