Understanding long-term interactions between climate, permafrost, and vegetation provides an essential context for interpreting current Arctic greening. Using 65 fossil pollen records from northern Siberia and a Random Forest model trained on a dataset of 835 modern pollen-climate assemblages, we quantitatively reconstructed mean temperature of the warmest month (Mtwa: mean July temperature) anomalies over the past 40 thousand years (ka) and assessed associated vegetation changes. During the Last Glacial Period, herbaceous taxa overwhelmingly dominated, and warming of similar to 1 degrees C during similar to 40-35 cal ka BP was insufficient to deepen the active layer beyond the threshold required for tree establishment, leaving woody cover minimal. In the early Holocene, sustained warming of nearly 2 degrees C triggered permafrost degradation and active-layer thickening, enabling forest expansion, although tree taxa lagged shrubs by several millennia. These results reveal a clear threshold effect in vegetation-permafrost interactions and show that only sustained warming can overcome permafrost constraints. By providing quantitative temperature estimates, our reconstruction offers critical benchmarks for predicting how ongoing Arctic warming may transform vegetation patterns and permafrost stability.