Abstract: Enhancing agricultural production efficiency in major grain-producing regions is crucial for ensuring national food security. Taking 102 counties in Henan Province, a core grain-producing area, as research units, this paper constructs a multi-input and multi-output agricultural production efficiency index system that incorporates multidimensional environmental variables and utilize a three-stage DEA model, spatial autocorrelation analysis, and multi-scale geographically weighted regression (MGWR) to uncover the spatiotemporal evolution characteristics and influencing factors of agricultural production efficiency in 2005, 2015, and 2022. Results demonstrate: 1) The comprehensive efficiency, pure technical efficiency, and scale efficiency of agricultural production in counties of Henan Province show an upward trend, with a notable rise in the number of counties achieving optimal efficiency; 2) The spatiotemporal evolution of agricultural production efficiency exhibits distinct spatial heterogeneity. Comprehensive efficiency and scale efficiency follow a stepped distribution pattern—higher in the eastern plains and lower in the western mountainous and hilly regions—while pure technical efficiency forms low-value clusters in transitional zones between plains and hilly areas; 3) Spatial autocorrelation analysis indicates that comprehensive efficiency, pure technical efficiency, and scale efficiency all display convergent clustering patterns (“high-high” and “low-low” agglomerations); And 4) the MGWR model highlights dynamic changes and spatial variability in the impacts of influencing factors. Industrial structure remains the dominant factor affecting agricultural production efficiency, although its influence has gradually weakened over time. This study provides theoretical and empirical support for optimizing resource allocation and formulating differentiated policies to enhance agricultural production efficiency in major grain-producing regions.