In this article, we present a molecular-level investigation of chemical dissociation in nitrogen at high temperature based on an accurate ab initio potential energy surface. For the first time in literature, both N-N2 and N2-N2 processes are concurrently simulated in an evolving gas system. The DMS simulations indicate that the presence of nitrogen radicals significantly affects the dissociation rate of molecular nitrogen and are consistent with experiments,1 which showed that atoms are more efficient at dissociating N2 than other nitrogen molecules. The analysis of normalized distributions of pre-collision states for dissociated N2 molecules reveals a strong bias to dissociation from high vibrational energy states, particularly at low temperature (10,000 K), similarly to N2-N2 only systems. However, the characteristic depletion of high vibrational energy states is less marked, thus suggesting that a mechanism is present that repopulates such levels at a faster rate than when only N2-N2 collisions occur. A possible explanation for such a behavior may be attributed to triatomic exchange processes, which were found to be very significant, particularly at low temperature (10,000 K). These processes appear to be quite effective at scrambling the internal energy states of the reactants. This, in turn, can facilitate transitions to high vibrational energy levels, which are more likely to dissociate, thus indirectly affecting the overall dissociation rate. Finally, discrepancies were found between Park’s two-temperature model2 predictions and the DMS results, particularly at the lowest temperature considered here (10,000 K). In general, composition histories obtained using Park’s rates show a slower dissociation process than the DMS results, over the whole temperature range. The comparison with Park’s model shows that the DMS technique is able to produce results, both macroscopic (energy and composition histories) and microscopic (distributions of reactive and nonreactive events), that can be used to develop and test simplified models for use in CFD or DSMC.