Personalized learning is an approach that tailors learning methods to meet the individual needs of learners in accordance with their unique characteristics and developmental potential. This study explores the application of large language models (LLMs) in personalized foreign language learning from the perspective of learners, evaluating the effects of different domestic and foreign models by comparing their performance across six dimensions. The results indicate that while domestic and foreign LLMs perform equally in addressing simple learning needs such as knowledge queries, foreign models excel in natural language generation and contextual meaning through accurately locating the context and appropriately responding to diverse interactive scenarios, whereas domestic ones perform clumsily in open-ended dialogues. Although the advantages of LLMs in promoting personalized experiences, lowering the threshold of learning, and promoting communication willingness will enhance the effect of personalized learning, potential issues also exist in relation to technological overdependence, false information, and academic misconduct. Accordingly, technology and humanities need to be integrated in future foreign language education, promoting the potential of LLMs while reducing possible risks.