Recent work has suggested that mid-IR wavelengths are optimal for estimating the mass-tolight ratios of stellar populations and hence the stellar masses of galaxies. We compare stellar masses deduced from spectral energy distribution (SED) models, fitted to multiwavelength optical-NIR photometry, to luminosities derived from WISE photometry in the W1 and W2 bands at 3.6 and 4.5 μmfor non-star forming galaxies. The SED-derived masses for a carefully selected sample of low-redshift (z≤0.15) passive galaxies agree with the prediction from stellar population synthesis models such thatM*/LW1 ≃0.6 for all such galaxies, independent of other stellar population parameters. The small scatter between masses predicted from the optical SED and from the WISE measurements implies that random errors (as opposed to systematic ones such as the use of different initial mass functions) are smaller than previous, deliberately conservative, estimates for the SED fits. This test is subtly different from simultaneously fitting at a wide range of optical and mid-IR wavelengths, which may just generate a compromised fit: we are directly checking that the best-fitting model to the optical data generates an SED whose M*/LW1 is also consistent with separate mid-IR data.We confirm that for passive low-redshift galaxies a fixed M*/LW1 = 0.65 can generate masses at least as accurate as those obtained from more complex methods. Going beyond the mean value, in agreement with expectations from the models, we see a modest change in M*/LW1 with SED fitted stellar population age but an insignificant one with metallicity.