We present co-seismic displacements of the Hyogo-ken Nanbu earthquake of January 17, 1995, detected by continuous GPS (Global Positioning System) observation, campaign type GPS survey and leveling. Continuous GPS observation gives a consistent pattern of displacements with those expected from a right lateral slip on a NE-SW trending vertical fault in far field: stations about 50 km east and west of the epicenter moved toward the epicenter by about 4 cm, while stations north and south moved away from the epicenter. By comparing with line lengths obtained by geodolite about 10 years ago, the campaign type GPS revealed most control points on Awaji Island moved to the southwest or south, which may be attributed to the movement of the Nojima Fault which cut the surface. On the other hand, control points northwest of the Rokko fault system moved toward the northeast and those on the other side moved slightly to the west, in and around Kobe. Leveling data revealed uplift of 19 cm on the northwestern side of the Suma Fault, a member of the Rokko fault system, and subsidence of 7 cm just east of this fault. Furthermore, uplift of about 5 cm was observed in the central part of Kobe City, and subsidence of 5 cm was detected east of Kobe. There is no significant gap in horizontal and vertical displacements around the northern extension of the Nojima Fault, which implies a complicated rupture process of this event. Leveling on the east coast of Awaji Island revealed a significant uplift of about 20 cm with slight subsidence at both edges of this uplift region during the past 20 years. By fitting the above geodetic data, we searched for an optimal set of parameters of a dislocation model. We assumed six nearly vertical faults trending NE-SW from Kobe to Awaji Island on the basis of aftershock distribution and focal mechanism. About 250 cm of the right lateral slip for the fault on Awaji Island is derived from large horizontal displacements near the Nojima Fault. The fault in Kobe may be divided into two segments with 100-200 cm slip by a slip-free zone which roughly corresponds to the cluster of aftershocks. The southern part of the Nojima Fault, segments near the Akashi Strait, and north of central Kobe may have significantly large thrust components of 100 cm. We also examined the possibility of buried faults beneath the zones of severe damage. Since this model cannot explain the observed geodetic data and the estimated slips are inconsistent with focal mechanism, these possible buried faults may not play a significant role, if any.