This paper presents the Weighted Penalty Model (WPM) for content balancing in computer adaptive testing. The WPM balances content properties as well as other non-statistical constraints, while simultaneously considering item information and the scarcity of items relative to content constraints, by assigning a penalty value to each eligible item in the item pool. Items with smaller penalty values are more desirable for selection. The WPM can be applied to both fixed/variable-length CAT once proper constraint codes and weights have been assigned. In this paper, applications of the WPM are presented using both empirical and simulated data. These analyses demonstrate the success of the approach in content balancing and in achieving targeted item exposure rates in the administration of CAT.