Abstract An interesting aspect of the El Niño–Southern Oscillation (ENSO) phenomenon is the asymmetry between its two phases. This paper evaluates the simulations of this property of ENSO by the Coupled Model Intercomparison Project phase 6 (CMIP6) models. Both the surface and subsurface signals of ENSO are examined for this purpose. The results show that the models still underestimate ENSO asymmetry as shown in the SST field, but do a better job in the subsurface. A much weaker negative feedback from the net surface heat flux during La Niña in the models is identified as a factor causing the degradation of the ENSO asymmetry at the surface. The simulated asymmetry in the subsurface is still weaker than the observations owing to a weaker dynamic coupling between the atmosphere and ocean. Consistent with the finding of a weaker dynamic coupling strength, the precipitation response to the SST changes is also found to be weaker in the models. The results underscore that a more objective assessment of the simulation of ENSO by climate models may have to involve the examination of the subsurface signals. Future improvements in simulating ENSO will likely require a better simulation of the surface heat flux feedback from the atmosphere as well as the dynamical coupling strength between the atmosphere and ocean. Significance Statement The ENSO phenomenon affects weather and climate worldwide. An interesting aspect of this phenomenon is the asymmetry between its two phases. Previous studies have reported a weaker asymmetry in the simulations by climate models. But these studies have focused on the ENSO asymmetry at the surface. Here by examining the ENSO asymmetry at the surface and the subsurface, we have found that ENSO asymmetry is better simulated in the subsurface than at the surface. We have also identified factors that are responsible for the degradation of the ENSO asymmetry at the surface as well as the remaining weakness in the subsurface, pointing out specific pathways to take to further improve ENSO simulations by coupled climate models.