With the rapid growth of e-commerce platforms in Vietnam, consumers increasingly rely on previous customer reviews and ratings, given the lack of physical product inspection before purchase. Recognizing the need for accurate sentiment classification in Vietnamese text, this paper proposes a combined model using PhoBERT and GPT optimized for the Vietnamese language to analyze sentiment in product reviews. By employing this combined model, the accuracy of product quality classification based on customer reviews is expected to enhance the decision-making process for consumers. We utilize a dataset from Kaggle for training and experimentation, comparing our results with those of similar studies. This paper aims to evaluate the performance of the combined PhoBERT and GPT models in sentiment classification, ultimately aiding consumers in making more informed purchasing decisions.