High-precision orbit determination of Global Positioning System (GPS) satellites is a key requirement for GPS-based precise geodetic measurements and precise low-Earth orbiter tracking. We explore different strategies for orbit determination with data from 1985 GPS field experiment. The most successful strategy uses multiday arcs for orbit determination and incorporates fine tuning of spacecraft solar pressure coefficients and stochastic station zenith troposphere delays using the GPS data. Average rms orbit repeatabilities for five of the GPS satellites are 1.0, 1.2, and 1.7 m in altitude, cross-track, and downtrack components, when two independent 5-day fits are compared. Orbits predicted up to 24 hours outside a 7-day arc show average rms component differences of 1.5--2.5 m when compared to independent solutions obtained wth a separate, nonoverlapping 5-day arc. For 246-km baseline, with 6-day arc carrier phase solutions for GPS orbits, baseline repeatability is 2 parts in 108 (0.4--0.6 cm) for east, north, and length components and 8 parts in 108 for the vertical component. For a 1314-km baseline with the same orbits, baseline repeatability is about 2 parts in 108 for the north component (2.5 cm) and 4 parts in 108 or better for east, length, and vertical components. When GPS carrier phase is combined with pseudorange, the 1314-km baseline repeatability improves further to 5 parts in 109 for the north (0.6 cm) and 2 parts in 108 for the other components (2--3 cm). ¿ American Geophysical Union 1987