In this report, the roasted Arabica coffees obtained from 4 Indonesian regions were analyzed with 1H NMR based-metabolomics. In total, 23 compounds were detected in the coffee 1H NMR spectra. Orthogonal projection to latent structure-discriminant analysis (OPLSDA) model successfully classified metabolites of the coffees based on their origins. S-plots of two-classes partial least square discriminant analysis (PLSDA) models successfully identified discriminant metabolites for every coffee. Chlorogenic acids, trigonelline, arabinoses were found as the discriminant compounds for Preanger-Java coffee. Lipids, acetic acid and lactic acid were discovered as the characteristic metabolites for Gayo-Sumatra coffee. γ-quinide was found as the most important marker for Bajawa-Flores coffee. Meanwhile, Toraja-Sulawesi coffee were characterized with a balance chemical composition indicating its well-balanced taste. The findings revealed the diversity of Indonesian Arabica coffees and shed more light on scientific information of Indonesian coffees.