An accurate Dynamic Traffic Assignment (DTA) model should capture real world traffic flow dynamics and predict ‘dynamic’ travel times. Traditional DTA models used simple traffic flow functions such as exit flow functions, delay functions, point queues, and deterministic physical queue models. Recently, simulation based models apply well accepted traffic flow theoretic models to simulate traffic flow. However, a significant number of papers over the last decade have adopted an approximation of LWR traffic flow model, the cell transmission model, for simulating traffic flow in a DTA model. This paper compares three models, namely, LWR, Payne and Aw-Rascle, models, for their suitability to be embedded in a DTA model. Model calibration and flow simulation is performed separately using two different speed–density relationships. Results showed the importance of choice of speed-density relationship in traffic flow simulation. Models were used to simulate traffic state at different discretization levels and it was observed that as discretization becomes finer, the models' accuracy increases. Finally, the models were applied to a two node, two link network to analyze their performance in a DTA framework. The higher-order models captured congestion dissipation better than LWR model which consistently underestimates congestion and travel time. © 2014, Springer Science+Business Media New York.