For describing the main field model at the 2000.0 epoch and the secular variation over the 2000-2005 timespan, three candidate models for the International Geomagnetic Reference Field (IGRF 2000) were proposed at the beginning of 1999, called in alphabetical order IPGP00 (proposed by IPGP), IZMI00 (proposed by IZMIRAN) and USUK00 (proposed by USGS/BGS). A fourth model, IGRF95 (the updated IGRF 1995), was suggested by the Working Group chairman. The modelling methods and the data used are presented by each team elsewhere in this special issue. This study is an attempt to test these models using the total field intensity provided by the Ørsted satellite, the only data available from that satellite at the time when the two tests describing here were done. The first test consists of evaluating the differences between the real and the synthetic data computed from the candidate models. The second test compares the capability of the candidate models to reduce the Backus effect, using a predictive dip-equator position and Ørsted data. Both tests show that the quality of the candidate models is far from being acceptable, and, therefore, a new candidate model for the main field, using vectorial Ørsted data, is required.