This paper provides a comprehensive survey of the current landscape of Large Model (LM) usage in complex game-playing scenarios, aiming to bridge the gap in systematic reviews of LM capabilities and potential in this area. The authors systematically review the architectures of LM-based Agents (LMAs) for games, highlighting their commonalities, challenges, and insights. They also present future research directions to advance LM capabilities in games. The survey is structured around the typical agent workflow: perception, inference, and action, and discusses the challenges and solutions at each stage. Key challenges include hallucinations, error correction, generalizing to unseen tasks, and enhancing interpretability. The paper concludes by identifying future research directions, such as improving multi-modal perception, achieving greater authenticity in gaming experiences, integrating external tools, and excelling in real-time gaming environments. The authors hope to inspire more research in this area and provide a clear understanding of the field.This paper provides a comprehensive survey of the current landscape of Large Model (LM) usage in complex game-playing scenarios, aiming to bridge the gap in systematic reviews of LM capabilities and potential in this area. The authors systematically review the architectures of LM-based Agents (LMAs) for games, highlighting their commonalities, challenges, and insights. They also present future research directions to advance LM capabilities in games. The survey is structured around the typical agent workflow: perception, inference, and action, and discusses the challenges and solutions at each stage. Key challenges include hallucinations, error correction, generalizing to unseen tasks, and enhancing interpretability. The paper concludes by identifying future research directions, such as improving multi-modal perception, achieving greater authenticity in gaming experiences, integrating external tools, and excelling in real-time gaming environments. The authors hope to inspire more research in this area and provide a clear understanding of the field.