Two feedlot trials were conducted to evaluate the 1996 NRC beef model under western Canadian conditions. In the first trial, 144 Charolais- (304.6 +/- 16.3 kg) and 144 Hereford- (295.1 +/- 20.8 kg) cross steers were used, whereas the second trial used 88 Angus- (289.7 +/- 15.0 kg), 88 Charolais- (299.8 +/- 17.9 kg), and 88 Hereford- (291.1 +/- 20.9 kg) cross steers. Diets were based on barley silage, rolled barley grain, canola meal, and cereal straw and were analyzed according to the 1996 NRC methodologies. Animal performance and environmental data were collected for 24 pens of steers per trial for the backgrounding and finishing periods. Levels 1 and 2 of the 1996 NRC model were used to generate predictions of DMI and ADG for each pen. Results showed that actual finishing DMI was accurately predicted for Trial 1 and for the combined trials but not for Trial 2. Predicted ADG was lower (P < 0.05) than actual ADG for all feeding periods except Level 1 of the Trial 1 finishing period. All ADG residuals were significant (P < 0.05), indicating inaccurate prediction of ADG in all feeding periods. The 1996 NRC model consistently predicted that protein was not limiting gain. Further investigations and model refinement regarding animal energy requirements under cold weather conditions and effects of limit feeding are required to increase the accuracy of the 1996 NRC model in predicting animal performance.