We analyse the redshift-space (z-space) distortions of QSO clustering in the 2dF QSO Redshift Survey (2QZ). To interpret the z-space correlation function, xi(sigma,pi), we require an accurate model for the QSO real-space correlation function, xi(r). Although a single power-law xi(r) model fits the projected correlation function (wp(sigma)) at small scales, it implies somewhat too shallow a slope for both wp(sigma) and the z-space correlation function, xi(s), at larger scales > 20 h^(-1) Mpc. Motivated by the form for xi(r) seen in the 2dF Galaxy Redshift Survey (2dFGRS) and in standard LCDM predictions, we use a double power-law model for xi(r) which gives a good fit to xi(s) and wp(sigma). The model is parametrized by a slope of gamma=1.45 for 1<r<10 h^(-1) Mpc and gamma=2.30 for 10<r<40 h^(-1) Mpc. As found for 2dFGRS, the value of beta determined from the ratio of xi(s)/xi(r) depends sensitively on the form of xi(r) assumed. With our double power-law form for xi(r), we measure beta(z=1.4)=0.32(+0.09)(-0.11). Assuming the same model for xi(r) we then analyse the z-space distortions in the 2QZ xi(sigma,pi) and put constraints on the values of Omega m and beta(z=1.4), using an improved version of the method of Hoyle et al. The constraints we derive are Omega m=0.35(+0.19)(-0.13), beta(z=1.4)=0.50(+0.13)(-0.15), in agreement with our xi(s)/xi(r) results at the ~1 sigma level.