In this study, artificial neural networks were used to model the hot deformation behavior of Zr-2.5Nb-0.5Cu alloy, in the strain rate range of 10-3 to 10 s-1, temperature range of 650-1050°C and to a strain of 0.5. Strain, log strain rate and inverse of temperature were used as inputs and stress was taken as the output of the network. The feed-forward network used consisted of two hidden layers containing four and three neurons each with a log-sigmoid activation function and Levenberg-Marquardt training algorithm. The network was successfully trained across phase regimes (α + β) to β and across different deformation domains. This trained network could predict the flow stress better than a constitutive equation of the type ε̇=A sinh(α′σ)nexp(-Q/RT). © 2005 Elsevier B.V. All rights reserved.