Ensemble filter assimilation methods represent a fusion of the ensemble prediction and data assimilation problems. The past few years have seen the development of a number of related ensemble filtering methods that are more or less closely related to traditional Kalman filters. One such filter, the ensemble adjustment Kalman filter is described as an approximation to solutions of the nonlinear Bayesian filtering problem. Results from a variety of perfect model observing system simulation experiments are used to demonstrate the wide-ranging capabilities of ensemble filters. Assimilations using a fully-parameterized global numerical weather prediction model demonstrate the filters' ability to work in very large systems. Assimilations of only observations of surface pressure in a global primitive equation model demonstrate an ability to address 'difficult' assimilation problems where the physically based relations between observations and state variables are not obvious. A brief comparison to the capabilities of 4D-variational assimilation in low order models includes an analysis of the computational cost of ensemble filter methods. Finally, a report on initial experiences from applying an ensemble filter in a Pacific basin version of the Modular Ocean Model (MOM) is presented.