Therapeutic antibodies have shown promise in treating diseases such as cancer, but selectivity remains a challenge, especially for solid tumors. To achieve a reasonable therapeutic window, antibodies must be selective for tumor cells while avoiding toxicity resulting from expression of tumor antigens on normal tissues. Bispecific antibodies can improve selectivity through avidity, which drives higher antibody binding on tumor cells expressing two membrane-bound targets compared to normal cells expressing only one of the targets, particularly when monovalent affinities are weak. However, it is often not clear what affinities will work best in the context of the expression levels of the targets of interest in vivo, and whether the expression levels of the chosen targets are in an optimal range. Moreover, predicting in vivo therapeutic windows from in vitro data may not be ideal due to the impact of PK effects. Mechanistic modeling can be used to optimize drug design as well as predict scenarios in which an avidity strategy is most likely to lead to a therapeutic window for a given set of targets. Here, we used a prebuilt model in Assess™, an interactive web-based model simulation software, to explore binding of a bispecific antibody across a range of drug affinities, avidities, and target expressions inside and outside a tumor.  We compared this with results from an in vitro model simulating binding to cells expressing one versus two targets.