Lifted diffusion flames are an interesting topic due to many reasons. Mainly, lifting the reaction zone provides explicit time for mixing and avoids, therefore, hot spots at near stoichiometric conditions. Hence, they promise low emissions and make them auspicious for industrial application. In comparison to lean premixed flames, which are promising in terms of emissions as well, they distinguish themselves in the nonexistence of the risk of flashback by concurrently nearly premixed flame conditions. From exploratory considerations they are an excellent case for the investigation of flame stability. However, especially this kind of flame is challenging for the reaction model due to its high turbulence and nearly premixed burning state. CFD is a powerful tool to get a clear insight in complex mechanisms, as it delivers detailed information of the flow field. Although the contribution of highly sophisticated models like LES is steadily growing in current research, fast models as RANS are most important. Solely they provide the feasibility of extensive parametric studies or the application in industrial design processes. Therefore, appropriate reaction models are needed. The applicability of two different reaction models for non-premixed flames to predict structure and stability of such flames has been investigated in this work. A stable confined diffusion flame produced by a double swirler airblast nozzle has been chosen as test case. Leaving the secondary air stream non-swirled creates a flame which stabilizes in a lifted state. The turbulent flamelet model as proposed by Peters in the early 90th basically models the impact of the turbulent strain rate on the diffusion flame. The local state of mixing is characterized by the mixture fraction, whereas the interaction of flame and turbulence is described by the mean scalar dissipation rate. The fact that the strain rate is the only non-equilibrium parameter describing the state of the reaction permits the use of detailed chemical mechanisms. The presumed jpdf model based on a 2-domain-1-step kinetic scheme has its focus on the interaction of mixing and reaction progress and uses a presumed shape for the joint probability density function. The reaction is characterized by a single variable describing the mixing state and one single additional variable, describing the state of reaction progress. In this paper assets and drawbacks of both models and their applicability to lifted flames have been discussed in detail. Furthermore, conclusions on the stability mechanism of a lifted swirl flame are taken.