The analysis of the PHEBUS FPT0 and FPT1 tests and some VERCORS tests performed under oxidising atmospheres with non irradiated (FPT0) and irradiated UO2 fuel evidenced that the fuel collapse temperature was well below the standard melting temperature of UO2 (3120 K) and above all lower than the 'eutectic' temperature of the UO2-ZrO2 pseudo-binary phase diagram (2820 K). Different assumptions as the role played by the fission products on the fuel degradation mechanism or the interaction between the structural materials and the fuel rods in the PHEBUS tests were advanced to explain a such behaviour. The fuel oxidation was also considered but on the basis of the old experimental data of Latta et al. , it was not retained as a dominant process . Liquidus and solidus temperatures were recently re-measured in the UO2+x composition domain by Manara . The main difference with the Latta's data is that the Manara's transition temperatures were accurately determined using a self-crucible technique while the former data were obtained in a W crucible and then suspected of crucible contamination. According to this recent data, a new thermodynamic modelling of U-O phase diagram is presented and introduced in the European NUCLEA thermodynamic database for Corium Applications . At high temperatures (T> 2000 K), the available information is not very abundant. Two invariant reactions may occur, e.g. (1) U3O8 ó Gas + UO2+x and (2) Gas + UO2+x ó Liquid. The temperature of the second reaction which is of great interest for the fuel collapse temperature in oxidising conditions is linked on the one hand, to the values of the oxygen potentials in the hyperstoichiometric solid solution region, which are extrapolated from the low temperature data and on the other hand, to the experimental data concerning the (Liquid + UO2+x) diphasic equilibrium. These latter values are properly defined if the liquidus shape is precisely known in this composition field. The new set of parameters fitting the Manara's data is then presented and it shows that the temperature of the reaction (2) is located around 2700 K at atmospheric pressure. An important consequence of this new optimisation for safety applications is that a liquid phase may appear in the O-UO2-ZrO2 composition domain of the U-O-Zr phase diagram at 2600 K at atmospheric pressure (this temperature decreasing with increase of pressure, about 2500 K at 2 atm.). This temperature should be still decreased by 100 K, depending of the physical model considered for the Gibbs energy description of the (U,Zr)O2+x fluorite structure. These temperatures can be associated with the temperature at which the fuel assembly could lose its integrity in oxidising conditions and then with what was observed in some of the VERCORS tests where fuel collapse was detected in the temperature range of 2400-2600 K or in the PHEBUS tests where indications of early fuel collapse at 2500-2600 K were identified.