CMIP5 and CMIP6 models, and a high-resolution regional model, were evaluated by comparing historical simulations with observations in the northwest North Atlantic, a climate-sensitive and biologically productive ocean margin region. Many of the CMIP models performed poorly for biological properties. There is no clear link between model resolution and skill in the global models, but there is an overall improvement in performance in CMIP6 from CMIP5. The regional model performed best.