The aim of this work was the development of multivariate models able to determine the content of the main crocetin esters and picrocrocin from spectrophotometric data that could be used for routine quality control of saffron. These compounds were determined with HPLC in Spanish saffron, and their absorbance spectra from 190 to 700 nm were simultaneously monitored. Partial least-squares regression (PLSR) models have been obtained and applied to the determination of individual crocetin esters, to the sum of crocetin esters, and to picrocrocin. A modification of the Kennard-Stone algorithm was used to divide the pool of samples into calibration and validation subsets. The best predictions were obtained with the sum of crocetin esters model, followed by the model for cis-crocetin (beta- D-glucosyl)-(beta- D-gentiobiosyl) ester, trans-crocetin di-(beta- D-gentiobiosyl) ester, and trans-crocetin (beta- D-glucosyl)-(beta- D-gentiobiosyl) ester, whereas the worst predictions were found with the picrocrocin and trans-crocetin (beta- D-gentiobiosyl) ester models. These models may enhance quality control in saffron enterprises.