The GRand Adaptive Sampling Experiment (GRASE) is a collaborative sea experiment that occurs in the Gulf of Mexico from April to September 2025. We employ our MIT-MSEAS data-assimilative Primitive-Equation (PE) submesoscale-to-regional-scale ocean-modeling system for real-time deterministic and probability forecasts of ocean fields and derived quantities. Our specific objectives include (i) multi-resolution ensemble forecasts with initial conditions downscaled from multiple models and implicit 2-way nesting, (ii) mutual information forecasts for predictability studies, (iii) optimal adaptive sampling guidance for multiple sea sensing platforms, and (iv) reachability forecasts for underwater vehicles. We illustrate a selection of our real-time results and products.