The surge in interest towards Vietnamese language education has led to the Vietnamese government establishing the Vietnamese Language Proficiency Framework (NLTV), a six-level scale akin to the CEFR for English. To support this initiative, we embarked on an innovative project to develop an automatic evaluation and classification model for written Vietnamese. Leveraging the powerful BERT model, combined with sophisticated assessments of phrase complexity, our approach aims to precisely categorize writings into the six NLTV proficiency levels. This cutting-edge NLP model not only promises to revolutionize Vietnamese language assessment but also sets a new standard for educational evaluation tools.