The Cornell Net Carbohydrate and Protein System (CNCPS) has equations for predicting nutrient requirements, feed intake, and feed utilization over wide variations in cattle (frame size, body condition, and stage of growth), feed carbohydrate and protein fractions and their digestion and passage rates, and environmental conditions. Independent data were used to validate the ability of the CNCPS to predict responses compared to National Research Council (NRC) systems. With DMI in steers, the CNCPS had a 12% lower standard error of the Y estimate (Sy.x) and three percentage units less bias than the NRC system. For DMI in heifers, both systems had a similar Sy.x but the NRC had four percentage units less bias. With lactating dairy cows' DMI, the CNCPS had a 12% lower Sy.x. Observed NEm requirement averaged 5% under NRC and 6% under CNCPS predicted values at temperatures above 9 degrees C but were 18% over NRC and 9% under CNCPS at temperatures under 9 degrees C. Energy retained was predicted with an R2 of .80 and .95 and a bias of 8 and 4% for the NRC and CNCPS, respectively. Protein retained was predicted with an R2 of .75 and .85 with a bias of 0 and -1% for NRC and CNCPS, respectively. Biases due to frame size, implant, or NEg were small. Body condition scores predicted body fat percentage in dairy cows with an R2 of .93 and a Sy.x of 2.35% body fat. The CNCPS predicted metabolizable protein allowable ADG with a bias of 1.6% with a Sy.x of .07 kg compared to values of -30% and .10 kg, respectively for the NRC system.