The aim of this study was to analyze the robustness and diagnostic value of different compartment models for dynamic ¹⁸F-FET PET in recurrent high-grade glioma (HGG). Dynamic ¹⁸F-FET PET data of patients with recurrent WHO grade III (n:7) and WHO grade IV (n: 9) tumors undergoing re-irradiation with carbon ions were analyzed by voxelwise fitting of the time-activity curves with a simplified and an extended one-tissue compartment model (1TCM) and a two-tissue compartment model (2TCM), respectively. A simulation study was conducted to assess robustness and precision of the 2TCM. Parameter maps showed enhanced detail on tumor substructure. Neglecting the blood volume V_B in the 1TCM yields insufficient results. Parameter K₁ from both 1TCM and 2TCM showed correlation with overall patient survival after carbon ion irradiation (p = 0.043 and 0.036, respectively). The 2TCM yields realistic estimates for tumor blood volume, which was found to be significantly higher in WHO IV compared to WHO III (p = 0.031). Simulations on the 2TCM showed that K₁ yields good accuracy and robustness while k₂ showed lowest stability of all parameters. The 1TCM provides the best compromise between parameter stability and model accuracy; however application of the 2TCM is still feasible and provides a more accurate representation of tracer-kinetics at the cost of reduced robustness. Detailed tracer kinetic analysis of ¹⁸F-FET PET with compartment models holds valuable information on tumor substructures and provides additional diagnostic and prognostic value.