A benchmark activity on two-fluid simulations of high-pressure boiling upward flows in a pipe is performed by 12 participants using different MCFD (Multiphase Computational Fluid Dynamics) codes and closure relationships. More than 30 conditions from DEBORA experiment conducted by CEA are considered. Each case is characterised by the flow rate, inlet temperature, wall heat flux and outlet pressure. High-pressure Freon (R12) at 14 bar and 26 bar is boiled in a 19.2mm pipe heated over 3.5m. Flow rates range from 2000 kg m−2 s−1 to 5000 kg m−2 s−1 and exit quality x ranges from single-phase conditions to x=0.1 which leads to a peak void fraction of α=70%. In these high pressure conditions, bubbles remain small and there is no departure from the bubbly flow regime (François et al., 2011; Hösler, 1968). However, different kind of bubbly flows are observed: wall-peak, intermediate peak or core-peak, depending on the case considered. Measurements along the pipe radius near the end of the heated section are compared to code predictions. They include void fraction, bubble mean diameter, vapour velocity and liquid temperature. The benchmark covered two phases. In the first phase of the benchmark activities, experimental data were given to the participants, allowing to compare the simulation results and to develop, to select or to adjust the models in the CMFD codes. The second phase included blind cases where the participants could not compare to the measurements. In between the two phases, possible additional model adjustments or calibrations were performed. Overall, the benchmark involved very different closures and a wide range of models’ complexity was covered. Yet, it is extremely difficult to have a robust closure for all conditions considered, even knowing experimental measurements. The wall-to-core peak transition is not captured consistently by the models. The degree of subcooling and the void fraction level are also difficult to assess. We were not capable of showing superiority of some physical closures, even for part of the model. The interaction between mechanisms and their hierarchy are extremely difficult to understand. Although departure from nucleate boiling (DNB) was not considered in this benchmarking exercise, it is expected that DNB predictions at high-pressure conditions depend strongly on the near-wall flow, temperature, and void fraction distributions. Therefore, the suitability of the closures also limits the accuracy of DNB predictions. The benchmark also demonstrated that in order to progress further in models development and validation, it is compulsory to have new measurements that include simultaneously as many variables as possible (including liquid temperature, velocity, cross-correlations and wall temperature); also, a better knowledge of the local bubble sizes distributions is the key to discriminate performances of interfacial area modelling (IATE, MUSIG or iMUSIG models, considering for instance the possibility of two classes of bubbles having totally different behaviour regarding the lift force). Following this benchmark impulse, we hope that future activities will be engaged on high-pressure boiling water experiments with a continuation of models’ comparisons and development.