Output from 42 models contributing to the Phase 6 of Coupled Model Intercomparison Project (CMIP6) have been used to assess the quality of their Arctic sea ice simulations for the period 1979-2014. According to passive satellite observations, the rate of decline in the Arctic sea ice has at least doubled since around 1997. To examine the multidecadal trends in sea ice area (SIA) and volume (SIV), we separated the 36-year time series into two subperiods (P1: 1979-1996 and P2:1997-2014) and examined their trends and acceleration ratios (i.e., P2/P1). We found the CMIP6 SIA multi-model mean (MM) underestimates the observed trend in SIA by 33% during the early period and 16% during the later period. This raises a concern that models underestimating the observed accelerated decline in sea ice are likely to underestimate the future projections of decline and the timing of a seasonally ice-free Arctic. Furthermore, the strong SIA rate of decline observed over P2 is mainly the result of the large September accelerated rate of decline between P1 and P2. In contrast, such a strong seasonal distinction is not shown in the CMIP6 SIA MM. Instead, the relative magnitude of the accelerated trend is essentially the same for March, September, and all other months. This points to some specific processes which are not realistically represented in CMIP6. A similar analysis was completed for SIV, but its interpretation is less conclusive due to the limited temporal coverage for SIV observational estimates and uncertainties in SIV trends based on reanalysis. We used the CMIP6 SIV MM as SIV reference and suggest that it may be a useful data set for future SIV studies. We also note that model internal variability plays a substantial role in the shorter subperiod trend analyses, as does the subperiod selection years.