Presented at the Global Ocean Salinity and the Water Cycle WorkshopThe Soil Moisture and Ocean Salinity (SMOS) mission monitors Sea Surface Salinity (SSS) from space since January 2010. This European Space Agency (ESA) Earth Explorer mission provided the first L-band radiometric observations of the Earth using interferometry. SMOS has demonstrated the feasibility of monitoring SSS and its variability from space with a precision of 0.15-0.3 (in regions free from radio frequency interferences and more than 1000km away from coasts). Some corrections however still need refinement such as the contamination by the land-sea emissivity gradient, the sun etcâ¦ (e.g. Boutin et al. 2016). Adjustments for systematic errors and radio frequency interferences have recently been successfully implemented on the level 2 and 3 SMOS operational chains. Empirical correction of the brightness temperatures in the ESA level 2 processing has indeed been made, adjusting better for land contamination. In the level 3 processing, a correction based on the self-consistency of SMOS SSS variations (Kolodziejczyk et al. 2016) has been implemented in Centre Aval de Traitement des DonnÃ©es SMOS (CATDS) for further correcting land contamination and systematic North-South seasonal errors. A particular attention has been put on preserving SSS natural variability in coastal areas, which was not well resolved by previous versions. These level 2 and 3 corrections greatly improve the SMOS-retrieved SSS. Systematic errors are reduced by more than 1 locally. The rms difference between the SMOS and Argo-derived SSS field (ISAS products; Gaillard et al. 2016) is reduced by more than 0.1 within 800km of the coast globally. The validation of this new retrieval method is however challenging due to the high natural SSS variability in coastal areas. We present a summary of the qualities and remaining flaws of these new retrievals, based on comparisons with SSS derived from SMAP (Soil Moisture Active/Passive), ISAS, ships of opportunity, surface drifters and TAO moorings.