With the increasing demand for productivity and quality in the forestry sector, near-infrared (NIR) spectroscopy is promising in the monitoring of wood properties, such as density. However, most predictive models are based on spectra acquired in wood at equilibrium moisture content using benchtop equipment. The objective of this study was to evaluate the performance of the NIR instruments in predicting the basic density of Eucalyptus grandis wood at different moisture contents. The wood chips were evaluated from saturated conditions (freshly felled) to hygroscopic equilibrium conditions using benchtop and portable NIR instruments. Principal component analysis (PCA) was performed to verify the behavior of spectral data, partial least squares discriminant analysis (PLS-DA) to classify density categories, and partial least squares regression (PLS-R) to develop predictive models. The moisture gradient was not the limiting factor for the statistical modeling. PCA discriminated 99.50% of the variation in the data, while the PLS-DA correctly categorized in the range of 0–94% the density classes. The models developed by PLS-R with the benchtop instrument showed a prediction coefficient (R²) ranging from 0.79 to 0.85 and those with the portable instrument ranged from 0.77 to 0.82; the ratios of prediction deviation (RPD) were 2.20 and 2.45, respectively. Thus, NIR spectroscopy has shown potential application in wood under saturated conditions, regardless of the type of instrument. In the industrial context, the use of a portable NIR instrument could streamline wood characterization without the need for drying and transporting samples to the laboratories.