This dataset provides corrected estimates of the natural land carbon sink (SLAND) for the period 1959-2024, accounting for the bias introduced by fixed pre-industrial land cover in Global Carbon Budget (GCB) simulations. The correction is based on outputs from seven Dynamic Global Vegetation Models (DGVMs) participating in the GCB2025, using a novel method that combines PFT-level net biome production (NBP) from the S2 simulation with transient land cover from the S3 simulation. The dataset includes: Global fluxes: SLAND values for each model before and after correction. Regional fluxes: SLAND for 10 global regions based on the RECCAP2 regional definitions. Gridded fluxes: SLAND (0.5° x 0.5°) estimates for each model, including a version incorporating S3 land cover. The Replaced Sinks and Sources can be calulcated as the difference between the two SLAND estimates. This dataset supports the manuscript "An improved approach to estimate the natural land carbon sink". The corrected SLAND estimate brings bottom-up land sink values into closer agreement with top-down atmospheric constraints and alters the Global Carbon Budget imbalance, with implications for understanding carbon-climate feedbacks and climate policy. ## Note: The sum of gridded fluxes will not match the global/regional values exactly due to the regridding of DGVM output to a common 0.5° grid. Also, models require their own land/sea masks to scale gridbox carbon flux values - which is taken into account in the provided global/regional values.