Seven isoenergetic semi-purified test diets containing graded levels of protein ranging from 20 to 50% were formulated using fish meal and casein as the protein sources. Test diets were fed to triplicate groups of Barbodes caldwelli juveniles with initial body weights of 1.26±0.02 g for eight weeks. The results indicated no significant effect of dietary protein levels on survival rate, relative weight of the viscera and relative weight of the liver in the juvenile fish. The weight gain and specific growth rate of the fish were found to be greater as dietary protein levels increased from 20 to 35%, but were not affected significantly as dietary protein level increased from 35 to 50%. Feed efficiencies did not differ significantly when fish were fed diets with protein levels from 30 to 50%, but were significantly higher than those of fish fed diets with protein levels of 20 and 25%. The protein efficiency ratio (PER) was negatively correlated with diet protein level (x) (PER=3.006−0.03251x, R=0.9366). There was no significant effect of dietary protein levels on carcass moisture, crude protein and ash. However, carcass lipid levels (L) decreased with an increase in dietary protein level (x) (L=8.2169−0.0458x, R=0.8551). There was no significant variation in hepatopancreas protease activity among the tests. Intestine protease activity and hepatopancreas amylase activity were increased to some extent, and then decreased as dietary protein levels continued to rise. The intestine amylase activity (IAA) of the juveniles was negatively correlated with dietary protein level (x) (IAA=84.625−0.9147x, R=0.8463). It was estimated that the suitable protein level for the B. caldwelli juvenile is 34% (the broken-line model was used to regress the relationship of the weight gain of the juvenile and dietary protein level).