Mismatch occurring during the fabrication of implant-supported dentures may induce stress to the peri-implant bone. The purpose of this study was to investigate the influence of two different alloys and the fabrication method on the marginal accuracy of cast partial dentures. Two laboratory implants were bonded into an aluminium block so that the distance between their longitudinal axes was 21 mm. Frameworks designed for screw-retained partial dentures were cast either with pure titanium (rematitan) or with a CoCr-alloy (remanium CD). Two groups of 10 frameworks were cast in a single piece. The first group was made of pure titanium, and the second group of a CoCr-alloy (remanium CD). A third group of 10 was cast in two pieces and then laser-welded onto a soldering model. This latter group was also made of the CoCr-alloy. All the frameworks were screwed to the original model with defined torque. Using light microscopy, marginal accuracy was determined by measuring vertical gaps at eight defined points around each implant. Titanium frameworks cast in a single piece demonstrated mean vertical gaps of 40 microm (s.d. = 11 microm) compared with 72 microm (s.d. = 40 microm) for CoCr-frameworks. These differences were not significant (U-test, P = 0.124) because of a considerable variation of the values for CoCr-frameworks (minimum: 8 microm and maximum: 216 microm). However, frameworks cast in two pieces and mated with a laser showed significantly better accuracy in comparison with the other experimental groups (mean: 17 microm +/- 6; P < 0.01). (i) The fit of implant-supported partial dentures cast with pure titanium in a single piece is preferable to that of those made with the CoCr-alloy and (ii) the highest accuracy can be achieved by using a two-piece casting technique combined with laser welding. Manufacturing the framework pieces separately and then welding them together provides the best marginal fit.