In its first 50 years, System Dynamics has evolved a rich set of modeling and analysis tools providing support for declarative, graphical specification of state equation models, with many features for analysis, calibration, and visualization.  Many System Dynamicists are now using Agent Based modeling packages, which offer very different feature sets from traditional System Dynamics tools.  The many differences between these two tool traditions have made many modelers uncertain about which toolset is most appropriate for their needs.  This paper analyzes differences between existing traditional System Dynamics and Agent Based tools, and discusses their impact on model quality attributes such as transparency, performance, and accuracy.  Our analysis finds that there is room for each tool tradition to benefit by adopting features of the other.  Looking forward, we argue that both traditions are likely to converge on support for features presently supported by just one tradition, including hybrid discrete/continuous time and states, declarative mathematical specification, modular composition, metadata and dimensional reasoning, and exploratory longitudinal visualization.  We believe that there will remain excellent reasons use a wide spectrum of model granularities, and argue competitive modeling packages will provide rich support for everything from extremely individual-level to aggregate models and multi-scale hybrids.