Since 2024, generative artificial intelligence represented by DeepSeek has achieved breakthroughs in Russian language training, providing technical conditions for the digital-intelligent transformation of Russian language education in higher education. This study examines the teaching practice of a Russian language course employing The Way to Russia 1 as its core textbook, systematically exploring implementation pathways and enabling mechanisms of large language models (LLMs) in collegiate Russian instruction under the“AI+HI”framework. Comprehensive analysis reveals that LLM-assisted Russian language education demonstrates three distinctive operational characteristics: multidimensionality, stratification and modality integration. They not only enable personalized modeling of disciplinary knowledge graphs, competency graphs, and problem graphs, but also possess technical potential for optimizing teaching lesson plans and assisting instructional activity design, achieving dual empowerment in both macro-level curriculum construction and micro-level instructional design. Through positive human-machine interaction adhering to the principles of “people-oriented” and “learning-oriented”, the integration of LLMs has effectively enhanced students’ comprehensive Russian language proficiency and intercultural communication skills, laying a crucial foundation for Chinese youth to better tell Chinese stories on the international stage.