This dataset contains the first public configuration of the CTSM land model designed for experimentation on CTSM hydrologic processes, including runoff. It contains the model implementation spanning a large-sample collection of 627 watersheds from the CAMELS dataset, also developed at NCAR. This archive contains the input, configuration and output files constructed to introduce a new model parameter estimation approach using AI-based model emulators with sequential optimization to calibrate and regionalize CTSM (and other models) for large-domain (national to global scale) implementation. This data resource (together with the public LSE method repository at NCAR) forms a valuable testbed for hydrologic modeling with CTSM will enable the community to explore and build on the new approach, and provides a benchmark for evaluating future modeling strategies for hydrology. It also contains a large sample (of up to 1000 parameter sets per basin) for assessment of CTSM hydrologic parameter sensitivities. The data are not grids but are mostly time-series of locations that enable the implementation of CTSM over 627 individual watersheds, treated individually as points.