Data from laboratory evaluations of seven fish bioenergetics models (BEMs) were used to investigate possible associations between BEM prediction error in relative growth rate (RGRerror) and levels of model input variables: mean daily food-consumption rate and fish body weight. Correlation between RGRerror and fish body weight was found in three BEMs applied under submaintenance feeding conditions. A strong correlation between RGRerror and mean daily consumption level was observed in all models over full consumption ranges; consumption level explained 70%–96% of variation in RGRerror. All BEMs underestimated (by 2- to 5-fold) growth at lower consumption levels and overestimated (by 2- to 3-fold) growth at higher consumption levels. RGRerror values associated with higher consumption levels were greater (up to 22 cal·g–1·day–1) than those at lower consumption levels (up to 10 cal·g–1·day–1). Correlation between consumption rate and RGRerror in all seven models indicates widespread systematic error among BEMs that likely arises from deficiencies in consumption-dependent model parameters. Results indicate that many BEMs are substantially inaccurate when predicting fish growth from higher feeding rates or estimating consumption from higher growth rates, even when higher consumption levels or growth episodes are of short duration. Findings obtained under submaintenance feeding conditions indicate that additional body-weight- and consumption-dependent terms should be added to BEM subequations for routine metabolism to account for metabolic reduction.