High-resolution soil moisture (SM) products are crucial for managing water in agricultural regions, irrigation scheduling, and land–atmosphere model simulations. Most satellite-based SM products have spatial resolutions in the tens of kilometers, which cannot satisfy the aforementioned applications, and there is an urgent need to disaggregate them to fine spatial resolutions. Currently, disaggregation method based on multi-source data fusion tend to be robust, but the following difficulties remain: 1) satellite products (microwave, optical/infrared) are somewhat missing in space and time which hinder the applications; 2) only one model-based product is used in the fusion process and different model-based products were not sufficiently evaluated. To overcome these problems, this work developed an SM disaggregation methodology. The best model-based SM product was first assessed using the triple collocation (TC) method, which allows us to obtain the weight of individual product. The model-based and microwave SM products were combined based on weight to produce daily seamless SM data at 25 km. Second, the iterative multi-temporal reconstruction and the ESTARFM method were combined to solve the invalid grids problem of optical/infrared product, thus obtaining 1 km daily seamless auxiliary data. Finally, daily seamless SM data at 1 km were obtained by geographically weighted regression (GWR) method. The methodology proposed in this study had a well-defined mechanism, and the resulting SM product achieved a satisfied accuracy (ubRMSE = 0.044 m3/m3), with characteristics of spatiotemporal continuity as well as fine spatial detail. The disaggregated SM product makes full use of model-based, microwave, and optical/infrared products, which is valuable in areas such as agricultural management policy, high-resolution hydro-meteorological simulations, and disaster monitoring.