In this paper, the problem of mixed additive/multiplicative model reduction for stable linear continuous-time systems is studied. To deal with nonsquare or nonminimum phase systems, the multiplicative error bound is constructed using spectral factorization technique. By virtue of the bounded real lemma and the projection lemma, a linear matrix inequality approach is proposed for mixed additive/multiplicative $H\u221e$ model reduction, which can be implemented by the well-known cone complementary linearization method. Finally, two numerical examples are provided to demonstrate the effectiveness and advantages of the obtained results.