An objective of the solar and space physics communities has been to predict the behavior of the interconnected physical systems that bring space weather to Earth. One approach is to use first-principles models that may predict behavior of the various space plasma regimes from the magnetized solar corona to Earth’s upper atmosphere. We focus on space weather forecasts in the thermosphere-ionosphere (T-I), with lead time based on the period following a solar eruption. There are generally 1-4 days lead time before the interplanetary coronal mass ejection (ICME) reaches the Earth’s magnetopause. Forecasting the behavior of the T-I with such multi-day lead times requires new ways of using and assessing first principles models, which are capable of predicting many details of the T-I response, including the time history of the global electron density distribution, neutral densities and neutral winds. All facets of the complex T-I system response must be predicted based on input solar and interplanetary parameters. Another influence on the forecast is the condition of the T-I at the time a forecast is produced (e.g. shortly after the CME eruption epoch). However, the role of such pre-conditioning is not well understood for lead times of a few days. To improve our understanding of these forecasts, we have submitted more than 120 multi-day simulation periods to NASA’s Community Coordinated Modeling Center, spanning three coupled T-I models. Approximately 40 T-I storms have been simulated, driven by solar wind and EUV parameters alone. We will present an analysis that characterizes how T-I models respond to the information content of the solar wind, mediated through climatological models of high latitude forcing, and the possible influence of pre-existing conditions. Smoothing across mesoscale variability is inevitable in this scenario. Analyzing the response across events and across models reveals critical information about the predictability of the T-I system as an ICME approaches.