Tea is one of the most widely consumed beverages in the world for its benefits to daily life and health. To discover the difference and correlation of chemical compositions in the three typical types of tea, a non-targeted metabolomics method was developed. After the optimization of extraction methods, gas chromatography-time-of-flight mass spectrometry and liquid chromatography-quadrupole time-of-flight mass spectrometry were applied for metabolomics analysis, 1812 and 2608 features were obtained, respectively. By comparing with the known compounds in public and/or commercial databases, 173 compounds were tentatively identified, and 109 of them were experimentally confirmed by standards. Totally, 33 tea samples including 12, 12 and 9 samples of green, oolong and black tea, respectively, were analyzed by using the above two methods. Multivariate analysis, Mann-Whitney U test and hierarchical cluster analysis were used to find and visualize the differential components in the three types of tea. Finally, 90 compounds, which contain catechins, amino acids, organic acids, flavonol glycosides, alkaloids, carbohydrates, lipids, etc, were found with a significant difference among them. This study demonstrates the potentials and power of metabolomics methods to understand the chemical secrets of tea. This should help a lot to optimize the processes of agriculture, storage, preparation and consumption.