Beginning in 2013, China launched two phases (2013–2017 and 2018–2020) of clean air actions that have led to substantial reductions in PM_2.5 concentrations. However, improvement in PM_2.5 pollution was notably slowing down during Phase II. Here we quantify the efficacy and drivers of PM_2.5 improvement and evaluate the associated cost during 2013–2020 using an integrated framework that combines an emission inventory model, a chemical transport model and detailed cost information. We found that national population-weighted mean PM_2.5 concentrations decreased by 19.8 μg m⁻³ and 10.9 μg m⁻³ in the two phases, and the contribution of clean air policies in Phase II (2.3 μg m⁻³ yr⁻¹) was considerably lower than that of Phase I (4.5 μg m⁻³ yr⁻¹), after excluding the impacts from meteorological condition changes and COVID-19 lockdowns. Enhanced structure transitions and targeted volatile organic compounds and NH₃ reduction measures have successfully reduced emissions in Phase II, but measures focusing on the end-of-pipe control were less effective after 2017. From 2013 to 2020, PM_2.5 abatement became increasingly challenging, with the average cost of reducing one unit of PM_2.5 concentration in Phase II twice that of Phase I. Our results suggest there is a need for strengthened, well-balanced, emission control strategies for multi-pollutants.