The sensitivity of Sargassum transport predictions to ocean circulation model resolution remains poorly quantified in the Mexican Caribbean, despite its relevance for operational forecasting systems. Here, we assess how horizontal resolution (5 km vs. 1 km) of simulations with the Regional Ocean Modeling System (ROMS) modulates Lagrangian particle trajectories and subsequent coastal landings. Four 10-day forecast case studies, including forcing conditions for different atmospheric and ocean dynamics, were analysed using particle tracking and finite-time Lyapunov exponent (FTLE) diagnostics to evaluate the effect of current spatial resolution on particle distribution and dispersion. Results reveal that resolution-dependent differences in landing percentages are strongly dependent upon the predominant circulation regime. Under typical conditions and circulation field, both resolutions produce comparable large-scale transport pathways and similar regional landing patterns. In contrast, under extreme event conditions (i.e., tropical cyclone), the 1-km simulation resolves wind-driven submesoscale structures that disrupt dominant advective transport, leading to substantial shifts in coastal accumulation patterns. The most significant differences in particles’ landings occurred in regions of complex coastal geometry, where shoreline curvature and island systems enhance sensitivity to small-scale circulation features. Despite these differences, both model resolutions identified recurrent accumulation hotspots near the Sian Ka’an Reserve, Banco Chinchorro, and Cozumel. These findings suggest that ocean circulation models with a 5 km spatial resolution are adequate for short-term regional forecasting of Sargassum beach landings, but higher-resolution simulations become critical under extreme events and in morphologically complex coastal environments in the Mexican Caribbean. More broadly, this study highlights the state-dependent nature of resolution requirements in Lagrangian coastal transport modelling and provides guidance for balancing computational cost and predictive skill in operational Sargassum forecast systems.