Background: Physical activity (PA) brings many benefits to pregnant women and fetuses; however, the majority of pregnant women do not participate actively in PA during pregnancy. Objectives: This study aimed to: (1) assess the utility of Multi-Theory Model (MTM) to explain the intentions of PA behavior in Chinese pregnant women; (2) analyze the predictors in initiating and maintaining PA behavior based on MTM. Methods: A cross-sectional study including pregnant women was conducted from March to June 2022 at a university hospital in Hangzhou, Zhejiang Province, China. Participants completed measures that included a self-developed demographic questionnaire and a 29-item MTM questionnaire. Descriptive statistics and stepwise multiple regression were used to analyze the data. The reliability was assessed by Cronbach’s alpha and test-retest stability. The construct validity was evaluated by using exploratory factor (EFA) analysis and confirmatory factor analysis (CFA). Results: A total of 450 pregnant women participated in this study. The score of the magnitude of intention to initiate and maintain PA behavior during pregnancy was 2.30 (1.08) and 2.24 (1.09). The overall Cronbach’s alpha value was 0.857. A four-factor structure for initiation model and a three-factor structure for maintenance model were determined. Results of the CFA confirmed construct validity of subscales (initiation model: χ2 = 206.123, df = 140, p < 0.001, χ2/df = 1.472, RMSEA = 0.046, SRMR = 0.0432, GFI = 0.913, CFI = 0.982; maintenance model: χ2 = 49.742, df = 29, p < 0.001, χ2/df = 1.715, RMSEA = 0.057, SRMR = 0.0432, GFI = 0.958, CFI 0.985). The result of regression indicated that participatory dialogue (β = 0.030; p = 0.002), behavioral confidence (β = 0.128; p < 0.001), changes in physical environment (β = 0.041; p = 0.005), trimester (β = −0.192; p = 0.001), and Gestational Diabetes Mellitus (GDM) (β = 0.408; p < 0.001) explained 52.1% variance in initiating PA behavior. Emotional transformation (β = 0.197; p < 0.001), practice for change (β = 0.083; p = 0.001), changes in social environment (β = 0.063; p < 0.001), pre-pregnancy exercise habit (β = −0.251; p = 0.001), and GDM (β = 0.298; p = 0.003) were significantly associated with pregnant women’s intentions to maintain PA behavior and explained 49.1% variance. Conclusions: The constructs of MTM were effective in explaining the intention to initiate and maintain PA behavior among Chinese pregnant women.