Partial least squares regression (PLS) was used to predict flash point and cetane index of diesel using distillation curves (ASTM-D86). The low RMSEP values obtained, compared with other chemometric models based on spectrometric methods described in literature, and high correlation coefficients between reference and predicted values showed that PLS was efficient to determine flash point and cetane index. The model built contains diesel samples of different compositions, thus revealing the variety of fuel in the Brazilian market. Furthermore, the proposed method has two advantages: low cost and easy implementation, as it applies the results of a routine test to evaluate the quality of diesel. (C) 2012 Elsevier Ltd. All rights reserved.