The objective of this study is to explore the differences in students' perceptions of grammar learning in online and offline classes at Kresna Kampung Inggris, as well as how these perceptions can potentially be utilized to optimize learning methods. 60 respondents-30 from the online class and 30 from the offline class-who had completed the Grammar 1 program participated in the descriptive-comparative quantitative methodology. A Likert scale questionnaire was adopted to collect data on six perception indicators, including course structure and organization, student interaction, instructor presence, student engagement, student satisfaction, and perceived learning. All of the indicators in the online class were classified as very positive, according to the results of the descriptive statistical analysis, whereas the offline class acquired positive categories for five indicators and very positive for the instructor presence indicator. As indicated by the results of the normality (Kolmogorov-Smirnov) and homogeneity (Levene's Test) tests, the data satisfied the requirements of parametric statistics. The Independent Sample t-test revealed a substantial distinction in students' perceptions of online and offline grammar classes, hence the alternative hypothesis (H1) was confirmed and the null hypothesis (H0) was disregarded. This result validates the unique characteristics of both methods that can be fully maximized. Instead than focusing on identifying the most appropriate method, this study aims to enhance the less successful aspects of each method while integrating its strengths. Therefore, the findings of this study argue for a more flexible grammar learning technique that is attentive to students' needs and varied learning environments.