Biogeochemical (BGC)-Argo floats observations are becoming a major data source for assimilation into and constraining of ocean biogeochemical models. An important prerequisite for a successful synthesis between models and observations is the characterization of the observational errors in BGC-Argo float data. The root-mean-square error and multiplicative and additive biases in qualitycontrolled data sets of oxygen, nitrate, and chlorophyll a concentrations collected with 17 BGC-Argo floats in the Mediterranean Sea between 2013 and 2017 are assessed using the triple collocation analysis. The analysis suggests that BGC-Argo float oxygen, nitrate and chlorophyll a data suffer from an additive bias of 2.9 ± 5.5 μmol/kg, 0.46 ± 0.07 μmol/kg, and −0.06 ± 0.02 mg/m 3 , respectively. The root-mean-square error is evaluated at 5.1 ± 0.8 μmol/kg, 0.25 ± 0.07 μmol/kg, and 0.03 ± 0.01 mg/m 3. Additional studies should determine whether these values are applicable to the global ocean. Plain Language Summary The Biogeochemical-Argo program is a network of ocean robots whose sensors monitor oxygen, nitrate, and chlorophyll a concentration information that is needed to detect decadal changes in biological carbon production, ocean acidification, ocean carbon uptake, and hypoxia in the world ocean. One of the goals of the Biogeochemical-Argo program is to incorporate these observations into ocean models to understand and forecast the changing state of the carbon cycle. The successful integration of the float data into numerical models, however, requires the specification of the observational errors. This study provides, for the first time, the biases and errors of the three cores variables of the Biogeochemical-Argo floats network: oxygen, nitrate, and chlorophyll a concentrations.