Abstract: The mixed layer (ML) is the ocean surface buffer layer, through which all buoyancy, momentum and gas exchanges with the atmosphere pass. It hosts a turbulence redistributing energy across-scales, either up- or down-scales, significantly contributing to the global ocean energy balance. Representation of ML turbulence, in numerical models and from observations, is therefore critical for investigating global ocean circulation and climate variability. We assess the characteristics of ML turbulence and its sensitivity to spatio-temporal resolution using a numerical simulation of the Drake Passage in winter — a region and period of intense meso and submesoscale turbulence. Here we show that the modeled winter ML turbulence is accurately inferred from hourly average numerical outputs. The ML is populated by geostrophic and ageostrophic submesoscale currents having a significant KE fraction in temporal scales of 1day-6hours up to 3 hours. The cross-scale KE fluxes shift from intensively upscale to more weakly downscale, where the KE becomes increasingly ageostrophic (KEageostrophic about 0.1-0.4 of KEgeostrophic, for spatial-scales O(>6) km)). This scale correspondance highlights the contribution to upscale (geostrophic currents) and downscale (coupled geostrophic-ageostrophic currents) KE fluxes. The characteristics of the ML turbulence are highly sensitive to temporal-scales. Daily-averaged outputs fail to represent the ML turbulence and 6 hourly-averaged outputs represent an intermediate turbulent regime, between interior quasi-geostrophy theory and ML turbulence. Our results physically-inform the need for high spatio-temporal resolution (O(1) km and O(1) h) to accurately infer the ML turbulence. Our results also call for further analysis on the role of ML turbulence on the regional dynamics of the Drake Passage.