Rapid Evaporative Ionization Mass Spectrometry (REIMS) has the capability to rapidly perform tissue analysis without sample preparation, extractions or chromatography required. The study was conducted to evaluate REIMS as an efficient platform to identify carotenoids in yellow-fleshed potato germplasm (N = 60) from the Colorado Potato Breeding and Selection Program. The specific aim eventually is to improve selection efficiency and accelerate genetic gain in nutritional quality of potato cultivars. Phenotypic tuber flesh color (FC) rating (0 – 3), chroma values, and individual and total carotenoids data were collected, processed and combined with multivariate analyses to help in REIMS data interpretation. Results showed that orange-fleshed (FC 3) potato genotypes gave significantly higher overall carotenoid content (P <0.0001) compared to the white-fleshed (FC 0), yellow-fleshed (FC 1) and dark yellow-fleshed (FC 2) genotypes. Zeaxanthin was the major carotenoid detected among the 60 selections/cultivars evaluated. The association between tuber flesh chroma and carotenoid content was analyzed. Results from Pearson correlation analysis revealed positive correlations overall. The correlation coefficient values (r) for lutein vs. chroma (r = 0.56, P < 0.01), zeaxanthin vs. chroma (r = 0.60, P < 0.01) and total carotenoid vs. chroma (r = 0.63, P < 0.01) were considered moderate. A metabolite mass fingerprint for each replicated sample was collected via REIMS to build a data matrix and processed to test the fit with prediction models. Multivariate methods of analysis (MVA) of principal component analysis (PCA), partial least square (PLS) and orthogonal-PLS (OPLS) were created to determine any sample differentiation among the yellow germplasm. FC rating data (0 – 3) were integrated to MVA models as a covariate. Rep 3 samples were excluded in all MVA analyses due to high presence of noise in the raw data. PCA of Reps 1 and 2 (n = 95) showed a predictive power of 48.4% (Q2). No apparent trends or separations based on flesh color was observed in the PCA model. PLS and OPLS supervised models illustrated better differentiation among sample components. OPLS model (n = 71) of high carotenoids (FC 3) vs. low carotenoids (FC 1 & 0) with a predictive power of 56.1% (Q2) was considered the best model due to clear separations of high vs. low carotenoid samples. Loadings and variable importance score (VIP) data were also analyzed to rank metabolite masses that contributed to differentiation of samples, detecting mostly lipid class molecules. Precursor molecules of lutein and zeaxanthin were not detected from the REIMS analyses and carotenoid fragmentation products were most likely contributing to differentiation among samples. Further research is needed to verify identification of carotenoid fragmentation in REIMS as well as the use of more portable and cost-efficient devices.