Previously, we have developed a gradient elution system for pillar array columns, which achieved faster separation than isocratic elution. In this study, we validated gradient elution in microchip liquid chromatography (LC) and investigated the retention and bandwidth predictions of fluorescently labeled aliphatic amines in fast gradient elution chromatography using semi-empirical retention models. The retention times and peak widths under different gradient elution programs were predicted by three solvent strength models and compared with the experimental results. The relative errors of prediction for the retention times and peak widths were below 14 % and 12 %, respectively. The results showed that the solvent strength models could be utilized for predicting the retention times and peak widths under gradient elution in the microchip LC system and that the gradient elution program for pillar array columns worked efficiently. The prediction by the retention model promises to be a potential tool for essential compound identification in biological samples.