A brief review is presented of the recent progress and the current state of the data-based modeling of the magnetospheric magnetic field. Combining the wealth of the observational data with flexible and realistic models advances our understanding of the dynamics of Earth's magnetic environment. The empirical approach to the modeling not only makes it possible to quantitatively represent the variable magnetosphere, but helps derive from data valuable information on its response to variations in interplanetary conditions. The cornerstones of the empirical modeling are (1) large sets of magnetic field and plasma data, (2) mathematical methods, allowing one to flexibly represent the B-field on a global scale, and (3) parameterization of magnetospheric field sources by the solar wind state variables and ground activity indices. This paper overviews most recent accomplishments made along these lines and discusses ongoing efforts to improve the data-based models; in particular, replicating the highly variable configuration of the magnetotail, introducing more flexible and realistic ring current model, and taking into account the variable shape and size of the magnetopause. Of special importance is the need to develop an improved representation of the high-latitude magnetosphere, including the dynamical large-scale Birkeland currents. Most of these problems can now be tackled, owing to recently devised powerful mathematical techniques, and a large amount of data obtained during several decades of spaceflight.