Green innovation has now become an important component of high-quality development, but China’s provinces still face imbalances in green innovation development. Clarifying the current state of development and the causes of the imbalance in each province is urgently necessary. But only a single or overall indicator cannot well reflect the structural differences within each province. We used the multi-way efficiency analysis (MEA) method to analyze the overall and structural efficiency of green innovation in each province of China, thus overcoming the shortcomings of comprehensive indicators. In addition, based on the decomposition of regional differences, policy factors resulting in heterogeneity among provinces are analyzed using Geodetector. The research results reveal the diversity of green innovation systems, the severe symmetry in resource utilization, and the internal and external sources of regional differences. We categorize the provinces into four development models by combining the internal structural characteristics of green innovation efficiency, as a way to propose suitable green innovation policies for each province, to take into account, the different development environments of the provinces assessed. Our research has significant implications for effectively improving green innovation efficiency and guides the formulation of more precise policies.