Abstract Recent reports suggest that large magnetic field changes occur prior to, and during, large earthquakes. Two continuously operating proton magnetometers, LSBM and OCHM, at distances of 17.3 and 24.2 km, respectively, from the epicenter of the 28 June 1992 Mw 7.3 Landers earthquake, recorded data through the earthquake and its aftershocks. These two stations are part of a differentially connected array of proton magnetometers that has been operated along the San Andreas fault since 1976. The instruments have a sensitivity of 0.25 nT or better and transmit data every 10 min through the GOES satellite to the USGS headquarters in Menlo Park, California. Seismomagnetic offsets of −1.2 ± 0.6 and −0.7 ± 0.7 nT were observed at these sites. In comparison, offsets of −0.3 ± 0.2 and −1.3 ± 0.2 nT were observed during the 8 July 1986 ML 5.9 North Palm Springs earthquake, which occurred directly beneath the OCHM magnetometer site. The observations are generally consistent with seismomagnetic models of the earthquake, in which fault geometry and slip have the same from as that determined by either inversion of the seismic data or inversion of geodetically determined ground displacements produced by the earthquake. In these models, right-lateral rupture occurs on connected fault segments in a homogeneous medium with average magnetization of 2 A/m. The fault-slip distribution has roughly the same form as the observed surface rupture, and the total moment release is 1.1 × 1020 Nm. There is no indication of diffusion-like character to the magnetic field offsets that might indicate these effects result from fluid flow phenomena. It thus seems unlikely that these earthquake-generated offsets and those produced by the North Palm Springs earthquake were generated by electrokinetic effects. Also, there are no indications of enhanced low-frequency magnetic noise before the earthquake at frequencies below 0.001 Hz.