Arctic maritime activity has been on the rise, with projections suggesting that ordinary commercial vessels may soon navigate high Arctic waters extensively, at least seasonally. This highlights the pressing need for precise assessments of Arctic marine access. Although climate models’ projections of sea ice thickness and concentration are widely employed to estimate Arctic shipping potential, their ability to replicate historical Arctic navigability remains insufficiently evaluated. Here, we show that the CMIP6 multi-model mean (MMM) underestimates Arctic navigability, despite accurately capturing historical changes in summer navigable areas. Since the early 21st century, the MMM has notably underestimated the shipping season in many Arctic Ocean regions. A clear dipole-like pattern emerges for moderately ice-strengthened (e.g., Polar Class 6) ships, with the MMM underestimating the season in the East Siberian and Laptev Seas, while overestimating it in the Kara and Barents Seas. Moreover, trans-Arctic routes through the Northwest Passage are absent from MMM estimation, likely due to inaccuracies in sea ice thickness simulations. Our analysis further shows that refining sea ice concentration estimates has a greater impact on assessing the shipping season for ordinary vessels, while improvements in sea ice thickness considerably enhance assessments for ice-strengthened ships. Furthermore, better sea ice thickness estimates are crucial for accurately identifying the fastest trans-Arctic shipping routes for both vessel types. Overall, our findings suggest that the MMM likely underestimates Arctic marine access and provide a scientific basis for improving the accuracy of Arctic navigability assessments.

sea ice, Arctic shipping, climate models, climate projection