This study was conducted to evaluate the predictions of dry matter intake (DMI), average daily gain (ADG), and faecal nitrogen (N) excretion by the Cornell Net Carbohydrate and Protein System Version 6.1.26 (CNCPSv6) in China. A total of 71 bulls from two imported breeds, Limousin and Simmental, and three local breeds: Luxi, Jinnan and Qinchuan were selected in China. Data required by the CNCPSv6 model were collected, and model predictions were generated for animals of each breed. The regression equation between observed and predicted DMI for these cattle was: Y_OBS = 0.93X_CNCPS + 0.48 (R² = 0.94; P < 0.001), with an intercept not different from zero and a slope not different from unity. The proportion of deviation points lying within the range –0.4 to 0.4 kg · d^–1 was 90.1%. The regression equation between observed and predicted ADG was: Y_OBS = 1.07X_CNCPS – 0.05 (R² = 0.92; P < 0.001), with an intercept not different from zero and a slope not different from unity. About 78.9% of points fell within the range –0.1 to 0.1 kg/d for these cattle. Model-predicted faecal N excretion for the cattle breeds was close to the observed values. The regression equation between observed and predicted faecal N excretion was: Y_OBS = 1.04X_CNCPS – 1.48 (R² = 0.94; P < 0.001), with an intercept not different from zero and a slope not different from unity. About 73.3% of the points fell within −4 and 4 g per day. These results show that the CNCPSv6 model using actual feed fractions can give good predictions of DMI, ADG and faecal N excretion with different beef cattle breeds in China.