Coastal polynyas are hotspots of winter sea-ice production in the Arctic Ocean, yet they remain poorly represented in climate models. Using the regional coupled ice-ocean model OPA-neXtSIM evaluated against remote-sensing observations from 17 Arctic polynya regions, we estimate pan-Arctic polynya activity over 18 winters (2000–2018). We diagnose polynya frequency, area, and ice production using classification methods based on ice thickness and thin-ice growth rate. The model successfully reproduces the observed spatial and temporal variability of Arctic polynyas, aided by its realistic representation of landfast ice. Depending on the classification method, accumulated polynya ice production ranges from 453 to 1014 km³ per winter, while total polynya area ranges from 167 x 10³ to 270 x 10³km². Taken together, these results demonstrate the importance of coastal polynyas for Arctic sea-ice production. Despite covering only about 2 % of the ice-covered Arctic Ocean, they account for up to 11 % of total winter sea-ice production. We further show that polynya activity depend strongly on the classification method used. In particular, thickness-based methods tend to overestimate polynya activity near the marginal ice zone. Consequently, previous observational estimates of polynya ice production derived from thin-ice thickness retrievals may be biased high.