This study examines the technical efficiency (TE) differences among typical cropping systems of smallholder farmers in the purple-soiled hilly region of southwestern China. Household-, plot-, and crop- level data and community surveys were conducted to explore TE levels and determinants of typical cropping systems by using a translog stochastic frontier production function. Results indicate significant difference in TE and its determinants among cropping systems. The mean TEs of the rice cropping system (R), the rice-rape cropping system (RR), the rice-rape-potato cropping system (RRP), and the oil cropping system (O) are 0.86, 0.90, 0.84, and 0.85, respectively, which are over 1.17 times higher than those of the maize-sweet potato-other crop cropping system (MSO) and the maize-sweet potato-wheat cropping system (MSW) at 0.78 and 0.69, respectively. Moreover, Technical inefficiency (TIE) of different cropping systems is significantly affected by characteristics of the household as well as plot. However, the impact of land quality, mechanical cultivation conditions, crop structure, farming system, farm radius, household type, cultivated land area per capita, and annual household income per capital on TIE vary by cropping system. Additionally, output elasticity of land, labor, and capital, as a group, is greater than the one of agricultural machinery and irrigation. Finally, when household-owned effective agricultural labor is at full farming capacity, optimal plot sizes for the R, RR, RRP, MSO, MSW, and O cropping systems are 1.12 hm2, 0.35 hm2, 0.25 hm2, 2.82 hm2, 1.87 hm2, and 1.17 hm2, respectively.