Context. The Gaia astrophysical parameters inference system (Apsis) provides astrophysical parameter estimates for several to hundreds of millions of stars. Aims. We aim to benchmark Gaia DR3 Apsis. Methods. We compiled approximately 1500 bona fide single stars in the Hyades and Pleiades open clusters for validation of PARSEC isochrones, and for comparison with Apsis estimates. PARSEC stellar isochrones in the Gaia photometric system enable us to assign average ages and metallicities to the clusters, and mass, effective temperature, luminosity, and surface gravity to the individual stars. Results. Apsis does not recover the single-age, single-metallicity characteristic of the cluster populations. Ages assigned to cluster members seemingly follow the input template for Galactic populations, with earlier-type stars being systematically assigned younger ages than later-type stars. Cluster metallicities are underestimated by 0.10–0.2 dex. Effective temperature estimates are in general reliable. Surface gravity estimates reveal strong systematic errors for specific ranges of the Gaia BP  −  RP colours. Conclusions. We caution that Gaia DR3 Apsis estimates can be subject to significant systematic uncertainties. Some of the Apsis estimates, such as metallicity, might only be meaningful for statistical studies of the time-averaged Galactic stellar population, but are not recommended to be used for individual stars.