A reassessment of the Nb-Ni system is proposed taking into account new experimental data. According to recent crystallographic work, the μ phase has been modelled using three independent sublattices of variable composition. This has been made possible by including the site fraction in the different sublattices as optimized experimental data. A consistent set of parameters is proposed. A good agreement is obtained between calculated and experimental values for all kinds of data (thermodynamic, crystallographic and phase diagram data).