Arctic and subarctic regions, which contain more than 33% of the world’s soil organic carbon, are warming up to four times faster than the global average and have poorly understood and modeled biophysical, biogeochemical, and ecohydrological processes. To better understand and constrain land models, especially regarding permafrost thaw and subsequent feedbacks to the earth system, The Warming Permafrost Model Intercomparison Project (WrPMIP) has modeled experimental permafrost warming sites across the Pan-Arctic. Though observational data in high-latitude ecosystems are limited, especially during winter, The WrPMIP has identified key observed responses that state-of-the-art carbon cycle models should accurately capture. We suggest that some of the most important processes to properly simulate are those related to soil freeze-thaw dynamics because of their importance in the net carbon balance of these systems. Here we focus on results from the WrPMIP’s first set of simulations that subsampled gridded regional baseline and warming responses around five long-term experimental warming sites. These simulations mimic soil respiration responses from open-top chamber (summer 0.5℃ soil and 1℃ air warming) and snow fence (1.5-3.0℃ winter soil warming) experiments. We compared the observed and modeled magnitudes and temporal dynamics of these responses and developed model benchmarks and uncertainty estimates for CLM, ELM, ecosys, and other models. Initial results from CLM5 show simulated OTC chamber warming of 0.06 g C m^-2 d^-1 ℃^-1, which is less than half the response expected when aggregating experimental warming results across the Pan-Arctic. Ongoing work to improve data-model agreement, including hypothesis testing and model development, will target improvements to terrestrial carbon cycle models and their representation of high-latitude ecosystems.