Abstract The computational codes used in the evaluation of the European sodium fast reactor—safety measures assessment and research tools(ESFR-SMART) reactor performance and specifically their sodium boiling models are assessed using two KNS-37 loss of flow (LOF) experiments, i.e., L22 and L29 tests, where boiling onset and two-phase flow regime up to dry-out occurred. The well-equipped KNS-37 experimental facility provided very valuable information for understanding the physical phenomena occurring in a 37-pin subassembly under LOF conditions, as well as experimental data to be used for computational tools validation. NATOF-2D, SAS-SFR, TRACE, ASTEC-Na, CATHARE-2, CATHARE-3, and NEPTUNE_CFD codes have been used in this exercise in order to compare the various boiling models and conclude on the advantages and limitations of them based on the comparison against the experimental data. Beyond boiling onset, the various sodium two-phase flow approaches determine the ability of the code to correctly represent the rewetting and voiding phases as well as cladding dry-out onset. A simulation performed by a computational fluid dynamics (CFD) approach (NEPTUNE_CFD code) taking into account liquid–vapor interfaces by an interface-tracking method is also shown and compared with the others approaches. Conclusions on each code performance are presented where the improvements needed to solve the issues encountered are included. This paper provides a first step in the process to investigate the required evaluation of the sodium two-phase flow models able to assess the safety of new SFR core designs (e.g., low void cores) under accidental conditions such as unprotected loss of flow (ULOF) transients.