AbstractDistribution of frozen ground and active layer thickness in the Northern Hemisphere during the mid‐Holocene (MH) and differences with respect to the preindustrial (PI) were investigated here using the Coupled Model Intercomparison Project Phase 5 (CMIP5) models. Two typical diagnostic methods, respectively, based on soil temperature (Ts based; a direct method) and air temperature (Ta based; an indirect method) were employed to classify categories and extents of frozen ground. In relation to orbitally induced changes in climate and in turn freezing and thawing indices, the MH permafrost extent was 20.5% (1.8%) smaller than the PI, whereas seasonally frozen ground increased by 9.2% (0.8%) in the Northern Hemisphere according to the Ts‐based (Ta‐based) method. Active layer thickness became larger, but by ≤ 1.0 m in most of permafrost areas during the MH. Intermodel disagreement remains within areas of permafrost boundary by both the Ts‐based and Ta‐based results, with the former demonstrating less agreement among the CMIP5 models because of larger variation in land model abilities to represent permafrost processes. However, both the methods were able to reproduce the MH relatively degenerated permafrost and increased active layer thickness (although with smaller magnitudes) as observed in data reconstruction. Disparity between simulation and reconstruction was mainly found in the seasonally frozen ground regions at low to middle latitudes, where the reconstruction suggested a reduction of seasonally frozen ground extent to the north, whereas the simulation demonstrated a slightly expansion to the south for the MH compared to the PI.