Positron emission tomography (PET) allows non-invasive detection and mapping of tumor hypoxia. However, slow tracer kinetics and low resolution, results in limited tumor-to-normal tissue contrast and the risk of missing areas where hypoxic cells are intermixed with necrosis. The shape of tumor time activity curves (TACs), as deduced from dynamic scans, may allow further separation of tumors/tumor sub-volumes that are inseparable based on static scans. This study was designed to define the added value of dynamic scans. Material and methods. Three squamous cell carcinoma tumor models were grown in mice. Mice were injected with the 18F-labeled PET hypoxia-tracer fluoroazomycin arabinoside (FAZA) and the immunologically-detectable hypoxia-marker pimonidazole, and PET scanned dynamically for three to six hours. Subsequently, microregional tracer retention (autoradiography) and the distribution of pimonidazole-retaining cells (immunohistology) and necrosis were analyzed in tumor tissue sections. Dynamic PET data were analysed based on a two-compartment model with irreversible tracer binding generating estimates of the putative hypoxia surrogate markers k3 (tracer trapping rate constant) and Ki (influx rate constant from plasma into irreversible bound tracer). Results/Discussion. High tumor-to-reference tissue ratios and a strong linear correlation (R∼0.7 to 0.95) between density of hypoxic cells and FAZA concentration was observed three hours after tracer administration, suggesting that late time PET images provides an accurate measure of hypoxia against which kinetic model estimates can be validated. Tumor TACs varied widely (ranging from distinctly wash-out to accumulative type) among tumor types although pimonidazole-stainings revealed extensive hypoxia in all models. Kinetic analysis of tumor sub-volumes showed that k3 correlated poorly with late time FAZA retention regionally in two of the three tumor models. The influx rate constant Ki displayed far less variability and correlated strongly with late time FAZA retention (hypoxia) in two of three tumor models, whereas a non-consistent relationship was observed in the last tumor model. Our study demonstrates the potential usefulness of dynamic PET, but also that a simple two-compartment model may be inappropriate in some tumor models.