This paper provides a comprehensive survey of the current status and opportunities for Large Language Models (LLMs) in strategic reasoning, a sophisticated form of reasoning that involves understanding and predicting adversary actions in multi-agent settings while adjusting strategies accordingly. Strategic reasoning is characterized by its focus on dynamic and uncertain interactions among multiple agents, requiring a deep understanding of the environment and the ability to anticipate others' behaviors. The paper explores the scopes, applications, methodologies, and evaluation metrics related to strategic reasoning with LLMs, highlighting the emerging developments and interdisciplinary approaches enhancing their decision-making performance. It aims to systematize and clarify the scattered literature on this subject, providing insights into future research directions and potential improvements. LLMs have revolutionized artificial intelligence, particularly in reasoning tasks such as common sense question answering and mathematical problem-solving. Strategic reasoning, which involves choosing the best action in a multi-agent setting by considering others' likely actions and the impact of one's decisions, is a critical cognitive capability. The necessity of strategic reasoning with LLMs extends beyond academic curiosity to understanding and navigating complex physical and social worlds. LLMs' text generation capabilities facilitate a wider range of strategic applications, while their powerful contextual understanding enables them to grasp new scenarios quickly, extending the scope of AI-based strategic reasoning beyond previous limits. The paper categorizes strategic reasoning scenarios into societal simulation, economic simulation, game theory, and gaming. Each category showcases LLMs' versatility and depth in understanding and influencing multi-agent dynamics. For example, societal simulations model human behavior in complex contexts, economic simulations analyze market dynamics, game theory simulations test strategic reasoning in competitive and cooperative situations, and gaming simulations enhance strategic depth in interactive entertainment. To enhance LLMs' performance in strategic reasoning, various methods have been developed, including prompt engineering, modular enhanced agents, Theory of Mind, and the integration of imitation learning and reinforcement learning. These methods aim to improve LLMs' situational awareness, adaptability, and strategic thinking capabilities. Evaluating LLMs in strategic reasoning involves both quantitative and qualitative assessments. Quantitative evaluations measure outcomes through metrics like win rates and survival rates, while qualitative evaluations focus on understanding the underlying mechanics of strategic reasoning, such as deception, cooperation, and discernment. The paper discusses the challenges and opportunities presented by applying LLMs to strategic reasoning. It highlights the need for systematic and rigorous research to understand the scalability and limitations of LLMs in complex strategic environments. The absence of unified benchmarks is identified as a key challenge, and the urgent need for collaborative efforts to create recognized benchmarks is emphasized. The paper also explores the unique challenges and potential of strategic reasoning for LLMs, suggesting that while larger models can capture more complex patterns, strategic reasoning fundamentally involves understanding intentions and predicting future actions. In conclusion, the paper underscores the pivotal role of LLMs in strategic reasoning, showcasing their evolution and significant advantages in complex decision-making across various domains. Future efforts should focus on interdisciplinary collaborations to bridge theoretical advancements and practical applicationsThis paper provides a comprehensive survey of the current status and opportunities for Large Language Models (LLMs) in strategic reasoning, a sophisticated form of reasoning that involves understanding and predicting adversary actions in multi-agent settings while adjusting strategies accordingly. Strategic reasoning is characterized by its focus on dynamic and uncertain interactions among multiple agents, requiring a deep understanding of the environment and the ability to anticipate others' behaviors. The paper explores the scopes, applications, methodologies, and evaluation metrics related to strategic reasoning with LLMs, highlighting the emerging developments and interdisciplinary approaches enhancing their decision-making performance. It aims to systematize and clarify the scattered literature on this subject, providing insights into future research directions and potential improvements. LLMs have revolutionized artificial intelligence, particularly in reasoning tasks such as common sense question answering and mathematical problem-solving. Strategic reasoning, which involves choosing the best action in a multi-agent setting by considering others' likely actions and the impact of one's decisions, is a critical cognitive capability. The necessity of strategic reasoning with LLMs extends beyond academic curiosity to understanding and navigating complex physical and social worlds. LLMs' text generation capabilities facilitate a wider range of strategic applications, while their powerful contextual understanding enables them to grasp new scenarios quickly, extending the scope of AI-based strategic reasoning beyond previous limits. The paper categorizes strategic reasoning scenarios into societal simulation, economic simulation, game theory, and gaming. Each category showcases LLMs' versatility and depth in understanding and influencing multi-agent dynamics. For example, societal simulations model human behavior in complex contexts, economic simulations analyze market dynamics, game theory simulations test strategic reasoning in competitive and cooperative situations, and gaming simulations enhance strategic depth in interactive entertainment. To enhance LLMs' performance in strategic reasoning, various methods have been developed, including prompt engineering, modular enhanced agents, Theory of Mind, and the integration of imitation learning and reinforcement learning. These methods aim to improve LLMs' situational awareness, adaptability, and strategic thinking capabilities. Evaluating LLMs in strategic reasoning involves both quantitative and qualitative assessments. Quantitative evaluations measure outcomes through metrics like win rates and survival rates, while qualitative evaluations focus on understanding the underlying mechanics of strategic reasoning, such as deception, cooperation, and discernment. The paper discusses the challenges and opportunities presented by applying LLMs to strategic reasoning. It highlights the need for systematic and rigorous research to understand the scalability and limitations of LLMs in complex strategic environments. The absence of unified benchmarks is identified as a key challenge, and the urgent need for collaborative efforts to create recognized benchmarks is emphasized. The paper also explores the unique challenges and potential of strategic reasoning for LLMs, suggesting that while larger models can capture more complex patterns, strategic reasoning fundamentally involves understanding intentions and predicting future actions. In conclusion, the paper underscores the pivotal role of LLMs in strategic reasoning, showcasing their evolution and significant advantages in complex decision-making across various domains. Future efforts should focus on interdisciplinary collaborations to bridge theoretical advancements and practical applications