<p>Accurate prediction of heatwaves is critical and remains a key objective of the Sub-seasonal to Seasonal (S2S) Prediction project. However, focusing on extreme cases and neglecting false alarms might overestimate regional forecast skill. Therefore, it is essential to evaluate the overall performance of current S2S models in predicting heatwaves. Using long-term datasets from current multiple subseasonal forecast systems, this study assessed the forecast skill for heatwaves in the mid-high latitudes of the Northern Hemisphere. Analysis revealed that heatwave frequency, intensity, and duration can be predicted up to 2 weeks in advance at the hemispheric scale. The S2S models generally predict the occurrence of heatwave 1–2 weeks in advance and warm anomalies 3–4 weeks ahead. Two notable patterns were identified: regional dependence and a duration effect. The former is exemplified by heatwaves in Europe being more predictable than those in East Asia and North America, which is attributed to higher inherent predictability and stronger model performance. The latter reflects the fact that long-duration heatwaves are more predictable than short-duration ones within the first lead 2 weeks, primarily owing to atmospheric predictability constraints. This study provides a more nuanced understanding of the current capabilities of S2S heatwave forecasts.</p>