An alternative method to predict distillation temperatures (DTs) in Brazilian crude oil samples using H-1 nuclear magnetic resonance (H-1 NMR) spectroscopy in association with partial least squares (PLS) regression is proposed. From the acquisition of 111 NMR spectra of 74 crude oil samples along with their DTs obtained by gas chromatography, PLS models were built to predict DTs from 5 to 90% distilled. In calibration and prediction steps, coefficients of multiple determination (R-2) and root-mean-square errors (RMSEs) were monitored throughout the studied range. The root-mean-square error of prediction (RMSEP, %) was the most important modeling parameter monitored, showing a decreasing behavior across the studied range and present values lower than 5% for some models. The residual analygis through bias and non-parametric permutation tests were taken into account for each built model. Lastly, the curves of DTs estimated by chemometrics, named DTEC, were faced with simulated distillation (SimDis) for each prediction sample. The developed methodology by H-1 NMR PLS can be successfully applied to Brazilian and similar crude oil samples to determine DTs from 5 to 90% &distilled.