Path planning is a fundamental problem in robotics and autonomous navigation, requiring efficient routes from start to destination while avoiding obstacles. Traditional algorithms like A* are effective but suffer from computational and memory inefficiencies as the state space grows. Large Language Models (LLMs) excel in broad environmental analysis but struggle with detailed spatial reasoning, often leading to invalid or inefficient routes. This paper proposes LLM-A*, a new route planning method that combines the precise pathfinding capabilities of A* with the global reasoning capabilities of LLMs. LLM-A* leverages LLM-generated waypoints to guide the path searching process, significantly reducing computational and memory costs. The algorithm integrates a standard L2 distance-based heuristic with new heuristic values derived from these waypoints, addressing granularity issues in LLM-generated solutions. Extensive experiments show that LLM-A* outperforms A* in terms of computational and memory efficiency, achieving nearly linear scalability in large-scale environments. The method retains the versatility of A*, making it suitable for a wide range of pathfinding tasks. The paper also discusses the limitations of the approach, such as the potential for suboptimal paths, and suggests future improvements to enhance path optimality.Path planning is a fundamental problem in robotics and autonomous navigation, requiring efficient routes from start to destination while avoiding obstacles. Traditional algorithms like A* are effective but suffer from computational and memory inefficiencies as the state space grows. Large Language Models (LLMs) excel in broad environmental analysis but struggle with detailed spatial reasoning, often leading to invalid or inefficient routes. This paper proposes LLM-A*, a new route planning method that combines the precise pathfinding capabilities of A* with the global reasoning capabilities of LLMs. LLM-A* leverages LLM-generated waypoints to guide the path searching process, significantly reducing computational and memory costs. The algorithm integrates a standard L2 distance-based heuristic with new heuristic values derived from these waypoints, addressing granularity issues in LLM-generated solutions. Extensive experiments show that LLM-A* outperforms A* in terms of computational and memory efficiency, achieving nearly linear scalability in large-scale environments. The method retains the versatility of A*, making it suitable for a wide range of pathfinding tasks. The paper also discusses the limitations of the approach, such as the potential for suboptimal paths, and suggests future improvements to enhance path optimality.