The ac loss of superconducting composite depends strongly on coupling between superconducting filaments via the resistive matrix. The established technique for loss reduction using twisted filaments relies on the decoupling of the filaments below a critical coupling field Bc, which increases with the reduction of the twist pitch and the matrix conductivity. Although the concept of Bc may be clearly demonstrated using two infinite slabs of finite length, further details on its correlation with the filament/conductor geometry are not yet available. The main obstacle is due to the fact that any accurate analysis of such a problem must be carried out in 3d. In this paper, we describe the initial results from 3d modeling using Cedrat‘s Flux3D, for which a superconductor module for handling power-law E-J characteristics was developed. Using a simple model of two rectangular superconductors connected through a normal metal, we demonstrate the feasibility for quantitative modeling of their coupling behavior over a wide range of field sweep rates for different conductor geometries. Typical examples were given for cases not addressed by the existing approximate theory, as well as for the evolution of field profile for varying field sweep rate. ample of a very low resistivity in the matrix, it is also suggested that onset of coupling is strongly dependent on the shape of the conductor and individual filaments. Qualitatively speaking, conductors with a large aspect ratio have a greater tendency to couple due to a larger demagnetising effect, as confirmed by experimental studies. While the mechanism that promotes the coupling of the filaments through the resistive matrix in ac magnetic fields is well understood conceptually, quantitative calculation of the critical coupling field is only limited to the simplified scenario of infinite slabs. The main difficulty is that we are dealing with a 3d problem with conductors of finite length and induced currents flowing both along and perpendicular to the long axis. Fortunately 3d calculations of superconductors in magnetic fields are becoming a reality thank to the rapid enhancement in computing power to desktop computers and progress in the implementation of numerical techniques for modeling superconductors.