Abstract. To study ocean acidification and the carbon cycle, marine scientists often use two measured or modeled carbonate system variables to compute others. These carbonate chemistry calculations, based on well-known thermodynamic equilibria, are now available from seven public packages: CO2SYS, csys, seacarb, swco2, CO2calc, ODV, and mocsy. We compared results from these packages using common input data and the set of equilibrium constants recommended for best practices. All packages agree within ±0.00025 units for pH and ±0.5 μmol kg−1 for CO32−, and six packages agree within ±0.2 μatm for pCO2 in terms of zonal-mean surface values. In the remaining package (csys), the surface pCO2 variable is up to 1.4 μatm lower than in other packages, but that is because it is mislabeled. When compared to surface fCO2, it differs by less than 0.2 μatm. The csys deviations in fCO2, pH, and CO32− grow with depth but remain small. Another package (swco2) also diverges significantly but only in warm deep waters as found in the Mediterranean Sea. Discrepancies between packages derive largely from their code for the equilibrium constants. Analysis of the sensitivity of each computed variable to changes in each constant showed the expected dominance of K1 and K2, while also revealing comparable sensitivity to KB, e.g., with the AT–CT input pair. Best-practice formulations for K1 and K2 are implemented consistently among packages, except those in csys deviate slightly at depth (e.g., 0.5% larger values at 4000 db) due to its pressure corrections made on the total instead of the seawater pH scale. With more recent formulations for K1 and K2 designed to cover a wider range of salinities, packages disagree more, e.g., by 8 μatm in pCO2, 1 μmol kg−1 in CO32−, and 0.006 units in pH under typical surface conditions. These discrepancies stem from packages using different sets of coefficients for the corresponding salinity dependence of the new formulations. Although each set should be equally viable after simple conversions, we show they are fundamentally inconsistent. Despite general agreement between current packages, agreement was much worse with outdated versions, e.g., differences reached up to 2.5 μatm in pCO2, 1.4 μmol kg−1 in CO32−, and 0.007 units in pH for surface zonal means when using the best-practice constants.