We have developed a generic dynamic model of extension of the lithosphere, which predicts major element composition and volume of melt generated from initial extension to steady state seafloor spreading. Stokes equations for non-Newtonian flow are solved and the mantle melts by decompression. Strengthening of the mantle due to dehydration as melting progresses is included. The composition is then empirically related to depletion. Using a crystallisation algorithm, the predicted primary melt composition was compared with mean North Atlantic mid-ocean ridge basalt (MORB). At steady state, using half spreading rates from 10 to 20 mm yr^{- 1} and mantle potential temperatures of 1300 to 1325 °C we predict a major element composition that is within the variation in the mean of North Atlantic MORB. This model is applied to the Southeast Greenland margin, which has extensive coverage of seismic and ODP core data. These data have been interpreted to indicate an initial pulse of magmatism on rifting that rapidly decayed to leave oceanic crustal thickness of 8 to 11 km. This pattern of melt production can be recreated by introducing an initial hot layer of asthenosphere beneath the continental lithosphere and by having a period of fast spreading during early opening. The hot layer was convected through the melt region giving a pulse of high magnesian and low silica melt during the early rifting process. The predicted major element composition of primary melts generated are in close agreement with primary melts from the Southeast Greenland margin. The observed variations in major element composition are reproduced without a mantle source composition anomaly.