Context. In spite of the importance of massive O-type stars for astrophysics, their accurate masses and other fundamental properties are still a matter of debate. Determining them reliably is hampered by various factors (stellar winds and other forms of circumstellar matter), and the agreement of derived properties with the model predictions is far from satisfactory. Careful studies of O-type binaries, especially of those in stellar clusters, are therefore desirable. Aims. Having obtained new series of electronic spectra and UBV photometry of V1007 Sco, we analysed these data in an effort to check whether the observed properties of V1007 Sco indeed disagree with the prediction of stellar evolutionary models. We briefly analysed data for a few other binaries in NGC 6231, too. Methods. Spectral reductions were carried out with the MIDAS program, photometry reduced using the HEC22 program, the orbital elements were derived with the FOTEL program and the final solutions obtained with the program PHOEBE. Results. Our analysis led to an accurate determination of the apsidal advance,ω = (0.00884 ± 0.00012) deg d −1 , based on a simultaneous solution of all usable radial-velocity and photometric data. This implies an apsidal period of 111.5 years. It is also demonstrated that the orbital inclination must be close to 67• . We arrived at the following preliminary values for masses and radii: M 1 = (29.5 ± 0.4) M , M 2 = (30.1 ± 0.4) M , R 1 = (15.8 ± 0.7) R , and R 2 = (15.3 ± 0.5) R . These values clearly indicate a log g of about 3.5 [CGS], implying that the stars are giants and not supergiants, as the standard spectral classification criteria indicate.