Several gasoline properties (alcohol, benzene, aromatics, saturated and olefin contents, density, MON, RON and distillation temperatures) have been predicted simultaneously by spectroscopy in the region of the mid infrared (FTIR) and partial least squares regression (PLS). The methods used as reference have included ASTM and ABNT methods and the IROX equipment, based on spectroscopy in the region of the near infrared. Although the standard error of the prediction (RMSEP) has been the main considered parameter to select the best model of prediction for each property, the repeatability and the reproducibility have also been evaluated. In general, the properties determined by IROX as the reference method (all the constituents, MON and RON) have presented poor results, the worst one found being the olefins content (21,3 +- 2,4) and the best one MON (82,5 +- 0,5). In the case of the properties that had ASTM or ABNT methods as reference (alcohol content, density and distillation temperatures) the worst results were found for the alcohol content (23,9 +- 0,7) and the best one for the density (0,7556 +- 0,0025). However, even values of RMSEP found by PLS regression had been considered small, the correlation coefficients (R2) between the predicted values and reference values, for a set of samples independent of the calibration, have been shown to be below 0,70, for all the properties. This fact can be explained by the small variation in the values of some properties, as MON, that in all gasoline samples presented values between 81,4 and 83,8. FTIR-PLS method revealed promising as an alternative for gasoline analysis. This method could be improved with the use of a greater set of calibration samples and/or with the use of a more representative sample set, beyond the use of standard reference methods.