In recent years, the emergences of large-scale and high-frequency serious haze weather mainly caused by air pollution in China have aroused widespread concerns at home and abroad, and thus the evaluation of China’s air pollution efficiency has become a topic of interest. In this paper, taking 2016 as the study period, we evaluate air pollution efficiency for China’s 30 mainland provinces by using an improved context-dependent slack-based measure (SBM) approach considering air pollution as the undesirable output under the total-factor production framework. The results show that utilizing the context-dependent SBM-Undesirable approach to evaluate provincial air pollution efficiency in mainland China is applicable and effective. Compared to other DEA-based approaches, the approach employed in this study can formulate more achievable short-term air pollution reduction targets for each China’s inefficient province in air pollution. In addition to this, it can also help confirm optimal learning benchmarks and appropriate benchmark-learning pathways for China’s inefficient provinces to combat air pollution. Finally, the conclusions and some related policy recommendations for the improvement of air pollution efficiency and air pollution mitigation in China are put forward.