Mixed-layer depth (MLD) and the stratification beneath it are key to ocean–atmosphere interactions, influencing how the upper ocean responds to atmospheric forcing. Their variability remains sparsely observed, particularly at subdaily timescales. Here we present Morlenn, a dataset of MLD and maximum buoyancy (Brunt-Väisälä) frequency squared, N<sup>2</sup>max covering the North Atlantic and adjacent equatorial waters from 2011 to 2023, at 0.25° spacing and three-hourly intervals. The dataset is generated from ocean surface satellite observations, atmospheric fluxes and wave forcing history using a neural network trained on Argo individual observations of MLD and stratification. Two versions are provided, one using surface inputs alone and one additionally incorporating neighbouring Argo data. Evaluation combines the broad coverage of Argo with independent glider observations. Scored against Argo stations over 2011–2023, the version using neighbouring data reaches an MLD RMS error of 38 m, against 45 m for ARMOR3D and 56 m for GLORYS on the same stations. It characterises MLD and <em>N</em><sup>2</sup><sub>max</sub> errors across regions and depth ranges, while spectral and diurnal-cycle analyses assess the variability represented by the product. Validation of subsurface stratification is less extensive. Morlenn provides new estimates of surface ocean mixed-layer depth and stratification for investigating upper-ocean variability, with documented limitations for shallow layers, extreme depths and very low salinity. The dataset is available at <a href="https://doi.org/10.17882/120058">https://doi.org/10.17882/120058</a> (Gaultier et al., 2026).