FLUXCOM used machine learning to merge carbon flux measurements from FLUXNET eddy covariance towers with remote sensing and meteorological data to estimate gross primary productivity (GPP), terrestrial ecosystem respiration (TER) and net ecosystem exchange (NEE) and their uncertainties. The resulting database comprises more than 100 global gridded products in two setups: (1) using exclusively MODIS remote sensing data (RS) on a 0.0833° grid, and (2) using remote sensing and meteorological data (RS+METEO) at 0.5° resolution. Within each setup we use a full factorial design across multiple machine learning methods, climate forcing datasets (for RS-METEO) and two flux partitioning approaches. Here we provide these data on a monthly resolution and provide also FLUXCOM ensemble products. Please see Jung et al. 2020, BG and Tramontana et al. 2016, BG before using the data, and cite both papers in publications. These data are freely available on a CC4.0 BY license.