Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents
Abstract
We present Game-TARS, a generalist game agent trained with a unified, scalable action space anchored to human-aligned native keyboard-mouse inputs. Unlike API- or GUI-based approaches, this paradigm enables large-scale continual pre-training across heterogeneous domains, including OS, web, and simulation games. Game-TARS is pre-trained on over 500B tokens with diverse trajectories and multimodal data. Key techniques include a decaying continual loss to reduce causal confusion and an efficient Sparse-Thinking strategy that balances reasoning depth and inference cost. Experiments show that Game-TARS achieves about 2 times the success rate over the previous sota model on open-world Minecraft tasks, is close to the generality of fresh humans in unseen web 3d games, and outperforms GPT-5, Gemini-2.5-Pro, and Claude-4-Sonnet in FPS benchmarks. Scaling results on training-time and test-time confirm that the unified action space sustains improvements when scaled to cross-game and multimodal data. Our results demonstrate that simple, scalable action representations combined with large-scale pre-training provide a promising path toward generalist agents with broad computer-use abilities.
Keywords
Cite
@article{arxiv.2510.23691,
title = {Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents},
author = {Zihao Wang and Xujing Li and Yining Ye and Junjie Fang and Haoming Wang and Longxiang Liu and Shihao Liang and Junting Lu and Zhiyong Wu and Jiazhan Feng and Wanjun Zhong and Zili Li and Yu Wang and Yu Miao and Bo Zhou and Yuanfan Li and Hao Wang and Zhongkai Zhao and Faming Wu and Zhengxuan Jiang and Weihao Tan and Heyuan Yao and Shi Yan and Xiangyang Li and Yitao Liang and Yujia Qin and Guang Shi},
journal= {arXiv preprint arXiv:2510.23691},
year = {2025}
}