Baijiu is globally recognized as one of the most prestigious distilled spirits. The traditional blending of baijiu plays a pivotal role in developing its unique flavor profile and overall quality. The complex flavor matrix of baijiu comprises both volatile compounds and non-volatile compounds. Volatile compounds primarily define aromatic perception, and non-volatile compounds significantly contribute to palate structure and overall body integration. Researchers have studied more on volatile compounds in baijiu and lack of studies on non-volatile compounds. Notably, systematic comparative analyses of non-volatile compounds dynamics across baijiu have remained conspicuously absent. To address this knowledge gap, non-targeted metabolomics were employed to analyze the non-volatile compounds in various baijiu samples using ultra-high performance liquid chromatography combined with electrospray ionization–triple quadrupole linear ion trap–MS/MS (UPLC–ESI–Q TRAP–MS/MS), and a total of 861 non-volatile compounds were identified. Based on differential metabolite analysis, Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis, and machine learning, the differences between the base samples and the commercial baijiu were explored, in addition to 7 pathways and 23 metabolic pathway markers that were found to be relevant to the formation of baijiu metabolites. By bridging traditional artistry with systems metabolomics, this study not only advances our understanding of the traditional blending process but also offers novel molecular targets for optimizing quality control in modern baijiu production.