Green development (GD) has become a new model of sustainable development across the world. However, our knowledge of green development efficiency (GDE) in Gansu province is poor. In remedy, this study, based on the panel data of 12 major cities in Gansu from 2010 to 2017, employed the super-efficient Slack-based measure (SBM) to analyze and evaluate GDE from the input–output perspective. Furthermore, we analyzed the input redundancy and output deficiency of identified inefficient cities in 2017 and conducted spatial autocorrelation analysis of GDE of the cities under study. Results show differences in the GDE of the major cities in Gansu, with an average value of 0.985. Green development efficiency in Lanzhou, Qingyang, Jinchang, Jiuquan, and Tianshui was relatively higher than in other cities. Green development efficiency in Zhangye, Wuwei, Jiayuguan, Baiyin, Dingxi, Longnan, and Longnan was less than one due to their redundant labor and capital input and excessive pollutant emission output. The overall GDE in Gansu depicts “high east and low west” zones. Each city in Gansu needs to formulate targeted policies and regulations to improve resource utilization, innovation capacity, reduce pollutant emission, optimize the industrial structure, and promote inter-city cooperation to construct a sustainable green economy.