A precise estimation of sediment transport capacity (Tc) is key to establishing process‐based erosion models. However, few data are available for estimating transport capacity on steep slopes and test materials sorted at high coarse grain values of >2 mm. Colluvial deposits with loose, coarse material and steep slopes make up the packed materials underlying the collapsing walls in benggang, which collapse due to hydraulic pressure and gravity. The objectives of this study were to investigate how flow discharge and slope steepness affect Tc and to examine relationships between Tc and flow velocity, shear stress, stream power, and unit stream power for colluvial deposits found on steep slopes. A nonerodible rill flume of 4 m long and 0.12 m wide was used. Slope steepness values ranged from 18% to 84%, and unit flow discharge values ranged from 0.56 × 10−3 to 4.44 × 10−3 m2 s−1. Tc increased as a power function with flow discharge and slope steepness with a Nash–Sutcliffe model efficiency (NSE) value of 0.99, and the effects of flow discharge were stronger than those of slope steepness. Tc was overestimated for a colluvial deposit when the equations of the ANSWERS, Zhang et al. and Wu et al. models were considered and when Tc exceeded 5 kg m−1 s−1, as the slope steepness used in our study was much higher than those used (<47%) in the other models. Regression analyses show that Tc can be predicted from linear equations of flow velocity, stream power, and unit stream power, and Tc can be fit to shear stress with power function equation. Flow velocity optimizes to predict Tc with NSE = 0.97, and stream power and shear stress can also be successfully related to Tc (NSE = 0.91 and NSE = 0.81, respectively); however, unit stream power performs poorly (NSE = 0.67). These results provide a basis for establishing process‐based erosion models on steep colluvial slopes.