Several mesoscale eddy-admitting, data-assimilative numerical circulation models for the Gulf of Mexico were run in a retrospective nowcast or analysis mode for 2010 as part of the Gulf of Mexico Pilot Prediction Project (GOMEX-PPP). Functioning in the style of a model testbed, skill assessment was conducted by GOMEX-PPP for these models, relative to each other and various observations, using several metrics. The results shed light on the veracity of contemporary modeling systems, the limits to their predictability, and the design of effective model testbeds.