Creep behaviour of metal alloys for critical high temperature engineering applications is usually studied by uniaxial tensile creep tests. The stress and temperature dependence of steady-state or minimum creep rate observed in such tests is rationalised at present on the basis of either Norton or MBD equation. But, the stress exponent determined on such basis depend strongly and often irregularly on test temperature and the creep activation energy varies with stress or stress range. Hence, these parameters cannot be used for long-term predictions. Here, it is shown that these difficulties can be removed if a new creep model, which incorporates tensile strength, is used to rationalise the creep data. The stress exponent determined then does not depend on temperature although it depends on stress range, and the creep activation energy does not depend on stress. Thus, there is no relationship between stress exponent value and creep mechanism. Consequently, the new tensile creep model can be used in combination with Monkman-Grant relationship to predict the long-term tensile creep strengths and lifetimes at different temperatures using creep parameters determined from short-term tensile creep tests. Both the deficiencies of the Norton and MBD equations and the predictive quality of the new tensile creep model are demonstrated here using uniaxial tensile creep data and tensile strength data measured for two totally different types of creep resistant engineering alloys: 9Cr-1Mo steel and Ni base superalloy Ni-16Cr-8.5Co-3.5Al-3.5Ti-2.6W-1.8Mo-0.9Nb.