Work zones considerably influence the traffic flow quality on freeways. In work zones, the capacity can be significantly reduced due to narrower lanes, a reduced number of traffic lanes, and unfavorable roadway geometry. In order to evaluate the impact of work zones on congestion occurrence and resulting delays as well as to allow for a consideration of alternative work zone layouts in the planning process, models to estimate the capacity of work zones are required. The paper presents results from a comprehensive study of work zone capacity on German freeways. In the analysis, a large number of short-term work zones with temporary lane closures as well as long-time work zones with and without a reduction of the number of lanes were covered. Data from loop or radar detectors upstream of the work zones were analyzed. For the capacity estimation of long-term work zones, both deterministic and stochastic approaches were used. In contrast to long-term work zones, traffic flow data measured at short-term work zones include only one traffic breakdown observation in most cases. This breakdown often occurs immediately after the lane closure, which means that the pre-breakdown volume does not represent the capacity of the work zone. Hence, valid capacity estimates could only be determined for the post-breakdown congestion outflow. The estimated capacities were analyzed regarding the influence of the traffic, geometric and control conditions in the work zone. Relevant parameters are the number of lanes, the lane widths, the existence of divided lanes, the existence of a lane reduction as well as the share of commuters and heavy vehicles. For short-time work zones, also the side of the lane closure (left or right lane) has an influence on the capacity. As a result of the study, capacity estimation models for both short- and long-term work zones, which can be applied in work zone planning procedures, are provided.