With the rapid advancement of artificial intelligence technology, its applications in the education sector have become increasingly profound. As a vital bridge for cross-cultural communication, international Chinese language teaching faces urgent demands for innovative teaching methodologies and efficiency improvements, particularly in core vocabulary instruction. Traditional vocabulary teaching methods often struggle to address learners’ individualized needs, the complex semantic relationships between words, and their rich cultural connotations. This paper explores a novel intelligent teaching system for international Chinese vocabulary that integrates knowledge graphs with large language models. Leveraging the powerful natural language understanding, generation, and interaction capabilities of large language models, the system provides more dynamic and user-friendly interfaces for knowledge graphs. The study systematically outlines the technical foundations and core mechanisms of knowledge graph-large language model integration, designs a system architecture tailored for international Chinese vocabulary teaching, and details the implementation approaches for key functional modules, including personalized learning path planning, contextualized vocabulary analysis, intelligent Q&A, and assessment generation. Finally, the paper examines multi-dimensional evaluation methods for the system and analyzes current technical challenges along with future development directions.