Faba bean ingredients are attracting interest for their suitability in producing protein-rich plant-based foods. However, their sensory characteristics (e.g., bitterness) challenge consumer acceptance. This study explored variations in the metabolome and the links between metabolites and sensory attributes using UHPLC-qTOF-MS/MS analysis of faba bean flour, two protein concentrates, and protein isolate. Partial Least Squares regression identified metabolites contributing to sensory descriptors, and it was validated against the VirtuousMultiTaste platform. Genetic variation and processing methods contributed to the metabolite composition of faba bean ingredients. We annotated 115 compounds with choline and vicine having the highest relative abundance. Five clusters suggested cultivar-specificity and process-related differences. Several compounds were linked to bitterness and mouth-drying orosensation, including caprolactam, gingerglycolipid, lysine, and vicine. Some compounds were reported as potentially bitter for the first time. This study lays the foundation for further research on the bitterness of these compounds and receptor-level investigations for targeted flavor optimization.