The performance of eddy-resolving global ocean–sea ice models in simulating mesoscale eddies is evaluated using six eddy-resolving experiments forced by different atmospheric reanalysis products. Interestingly, eddy-resolving ocean general circulation models (OGCMs) tend to simulate more (less) energetic eddy-rich (eddy-poor) regions with a smaller (larger) spatial extent than satellite observation, which finally shows that larger (smaller) mesoscale energy intensity (EI) is simulated in the eddy-rich (eddy-poor) regions. Quantitatively, there is an approximately 27%–60% overestimation of EI in the eddy-rich regions, which are mainly located in the Kuroshio–Oyashio Extension, the Gulf Stream, and the Antarctic Circumpolar Currents regions, although the global mean EI is underestimated by 25%–45%. Apparently, the eddy kinetic energy in the eddy-poor region is underestimated. Further analyses based on coherent mesoscale eddy properties show that the overestimation in the eddy-rich regions is mainly attributed to mesoscale eddies’ intensity and is more prominent when mesoscale eddies are in their growth stage.