With the development of point source pollution control, Agricultural Non-point Source Pollution (AGNPS) issues have become increasingly prominent worldwide. Non-point source pollution is difficult to control because it comes from the everyday activities of many different people, such as fertilizing a lawn, using a pesticide, or constructing a road or building. Recently, the agricultural non-point source pollution control has become a hotspot in the water research. As a common tool used to reduce non-point source pollution, Best management practices (BMPs) have been widely adopted to improve water quality problems associated with agricultural nonpoint source pollution, however, there have been few realistic efforts to assess their effectiveness in reducing AGNPS pollution. The effectiveness of BMPs must be evaluated at various spatial and temporal scales before adoption. Models are more comprehensive that can reflect choice of mitigation at a widely range of scales and then to achieve the best cost-effectiveness selection and placement of BMPs for non-point source pollution control. In this paper, we review some models used to assess the effectiveness of BMPs for agricultural non-point source pollution control, including non-structure practices and structure practices. The conceptual models mostly used to evaluate the impact of source control measures, while the physically-based models used to evaluate the BMPs that through control the timing and location, the response time and the transport and transformation of pollutants. The lag time between adoption of management changes and the detection of measurable improvement in water quality in the target water body are extremely important for the BMPs estimation as well as the model evaluation and validation. Models can be served as an effective tool to identify timing and critical non-point source pollution areas for target actions at different spatial scales. Other issues of critical importance include minimizing pollution swapping and assessing the cost-effectiveness of the measures within multi-objectives, as well as the acceptance of the these measures by the stakeholders involved before performing an integrated assessment of BMPs implementation. These issues are all relevant and challenging for the implementation of water and environmental policies. For future research, approaches to deal with the inevitable lag time between implementation of management practices and water quality response lies in appropriately characterizing the watershed, considering lag time in selection, location, and monitoring of management measures including the selection of appropriate indicators and designing an effective monitoring programs to detect water quality response. Understanding of NPS model uncertainty has become a front edge topic, and future studies should focus on improvement of parameter calibration, optimization of data acquisition solutions, and uncertainty analysis. Regarding to the timing and location of measures, pollution and ecosystem service swapping, and optimization and placement of BMPs in watershed, the integration of NPS models with 3S technology (GPS, RS, GIS) should be proposed. Stakeholders may play important role in developing the mitigation plan and enhancing the communication, reciprocal understanding, trust and acceptance of modelling results.