A model of cross-bridges binding to actin in the weak binding A*M*ATP state is presented. The modeling was based on the x-ray diffraction patterns from the relaxed skinned rabbit psoas muscle fibers where ATP hydrolysis was inhibited by N-phenylmaleimide treatment (S. Xu, J. Gu, G. Melvin, L. C. Yu. 2002. Biophys. J. 82:2111-2122). Calculations included both the myosin filaments and the actin filaments of the muscle cells, and the binding to actin was assumed to be single headed. To achieve a good fit, considerable flexibility in the orientation of the myosin head and the position of the S1-S2 junction is necessary, such that the myosin head can bind to a nearby actin whereas the tail end was kept in the proximity of the helical track of the myosin filament. Hence, the best-fit model shows that the head binds to actin in a wide range of orientations, and the tail end deviates substantially from its lattice position in the radial direction (approximately 60 A). Surprisingly, the best fit model reveals that the detached head, whose location thus far has remained undetected, seems to be located close to the surface of the myosin filament. Another significant requirement of the best-fit model is that the binding site on actin is near the N terminus of the actin subunit, a position distinct from the putative rigor-binding site. The results support the idea that the essential role played by the weak binding states M*ATP <--> A*M*ATP for force generation lies in its flexibility, because the probability of attachment is greatly increased, compared with the weak binding M*ADP*P(i) <--> A*M*ADP*P(i) states.