Just because they are higher res, doesn't mean they are more accurate - I mean, they are spitting out a lot of mix/ZR and we know that's not likely going to happen. I'd bet they''ll show a more east track 24hrs from the event.
From Reddit:
The 12km NAM has a resolution of 12km and the 3km NAM has a resolution of 3km. But to give more specifics:
The NAM is a specific implementation of the Weather Research and Forecasting (WRF) model. WRF supports the use of "nesting": You have a larger, "parent" domain run at fairly low resolution, and one or more "nested" domains at higher resolution. The model runs a calculation for one timestep of the parent domain, then those conditions are interpolated down to the nest domains, they run the calculations on the high-resolution nests for a few timesteps, then interpolate that final result back up to the parent domain, calculate a timestep, and so on.
The main advantage of running the model this way (instead of just running at 3km resolution across the whole domain) is that we can put a lot more computational power into modeling phenomena over the areas we care about, rather than wasting a compute time by modeling the atmosphere at very high resolution over the open ocean (for example) where we don't care so much about the fine-scale details of the weather.
I believe the only difference aside from resolution is that the convective parameterization schemes may be modified or turned off in the 3km dom