The reduction in the quality, consumption, and digestibility of forage can cause a decrease in animal performance, resulting in losses to the rural producer. Thus, it is important to monitor these characteristics in forage plants to devise strategies or practices that optimize production systems. The aim of this study was to develop and validate prediction models using near-infrared spectroscopy (NIRS) to determine the chemical composition of Tifton 85 grass. Samples of green grass, its morphological structures (whole plant, leaf blade, stem + sheath, and senescent material) and hay, totaling 105 samples were used. Conventional chemical analysis was performed to determine the content of oven-dried samples (ODS), mineral matter (MM), crude protein (CP), neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL), cellulose (CEL), hemicellulose (HEM), and in vitro dry matter digestibility (IVDMD). Subsequently, all the samples were scanned using a Vis-NIR spectrometer to collect spectral data. Principal component analysis (PCA) was applied to the data set, and modified partial least squares was used to correlate reference values to spectral data. The coefficients of determination (R2) were 0.74, 0.85, 0.98, 0.75, 0.85, 0.71, 0.82, 0.77, and 0.93, and the ratio of performance deviations (RPD) obtained were 1.99, 2.71, 6.46, 2.05, 2.58, 3.84, 1.86, 2.35, 2.09, and 3.84 for ODS, MM, CP, NDF, ADF, ADL, CEL, HEM, and IVDMD, respectively. The prediction models obtained, in general, were considered to be of excellent quality, and demonstrated that the determination of the chemical composition of Tifton 85 grass can be performed using NIRS technology, replacing conventional analysis.