Three-dimensional magnetic modeling of the North Atlantic and northeast Pacific is performed at intermediate wavelengths using three models for the acquisition of a natural remanent magnetization. It is shown that a remanent magnetization which is dependent on the crustal age is the dominant source for the intermediate-wavelength pattern in both basins. However, a pure thermoremanent magnetization of layer 2 alone is insufficient to model the intensity and shape of the observed magnetic anomalies at satellite altitude. We conclude that the best fitting magnetization model for both basins is a combination of a chemical remanent magnetization in the altered upper crust and a thermoviscous remanent magnetization of the slowly cooling lower crust and upper mantle. The North Pacific requires a bulk magnetization which is 50% higher than that of the North Atlantic in order to fit the Magsat field. Geological processes associated with a faster spreading rate such as a faster hydrothermal alteration and the growth of a thicker gabbro layer at the expense of a weakly magnetized sheeted dike layer are plausible explanations for the higher North Pacific magnetization. The lineated positive magnetic anomaly observed over the North Atlantic spreading center is not well reproduced by our models. This anomaly is likely due to a highly magnetized body along or in the vicinity of the spreading center. This highly magnetic body could be an unstable serpentinized lens of crustal material younger than 20 m.y. which develops within the zone cooled by hydrothermal circulation.¿ 1997 American Geophysical Union