As recently reported, the metal–carbon eutectic points seem to be useful as practical reference points for calibrating thermocouples. The melting temperature of the cobalt–carbon eutectic point (Co–C, 1324◦C) is an ideal solution for a reference point being at approximately the middle between the freezing point of copper (1084.62◦C) and the melting point of palladium (1554.8◦C). In this study, a comparison of reference Co–C cells used for thermocouple calibration service was performed between the Technical Institute of Sweden (SP) and the National Metrology Institute of Japan (NMIJ) by means of two circulating Pt/Pd thermocouples and one circulating Co–C cell in order to investigate the reliability of thermocouple calibration at the Co–C eutectic point. The reference Co–C cell at SP was designed and constructed in the National Physical Laboratory (NPL), and provided to SP, while the reference Co–C cell at NMIJ as well as the circulating cell was designed and constructed at NMIJ. The melting temperature of the reference Co–C cells at SP and NMIJ were evaluated by using the calibrated radiation thermometers at NPL and NMIJ, respectively, prior to the circulation of the thermocouples and the cell between SP and NMIJ. The temperature difference between the reference cells of SP and NMIJ, as measured by means of the circulating thermocouples, was found to be within 0.33◦C, while that measured by means of the circulating cell was found to be within 0.22◦C. The temperature of reference cells at SP and NMIJ was found to agree within an expanded uncertainty (k=2) in spite of different temperature traceability chains. This means that the calibration of thermocouples at the Co–C eutectic point was sufficiently reliable.