Near-infrared IR reflectance spectroscopy (NIRS) was applied to the compositional analysis of oysters (Crassostrea gigas and Saccostrea glomerata). Homogenized meat samples of 332 oysters were scanned by NIRS, subsamples were analyzed chemically, and, by combining the sets of information, calibration models were developed to allow prediction of proximate composition (moisture, protein, glycogen, and fat). Predicted and actual (chemically measured) data in independent validation sample sets were compared using R² and the ratio of the SE of chemical data to the SE of NIRS prediction (RPD). For S. glomerata, models gave excellent prediction for all components (R² = 0.95–0.97, RPD = 2.7–5.5). Prediction within the C. gigas validation set was generally less precise, but still very good for all components (R² = 0.92–0.96, RPD = 2.7–4.8). With a smaller subset of samples (n = 48), prediction models were also developed for estimating concentration of polyunsaturated fatty acid and long-chain polyunsaturated fatty acid (R² = 0.94 and 0.93, respectively). The major advantages of the methodology are its speed—250—300 samples can be analyzed simultaneously for all components each day—and cost-effectiveness when a large number of samples (e.g., several hundred or more per year) are analyzed. Therefore, the method is ideally suited to applications requiring the rapid analysis of many individuals, such as selective breeding programs for which chemical compositional data can provide information on traits associated with oyster condition or quality.