Most simulation studies done to deeper understand high-intensity welding arcs address axi-symmetric configurations and use the electric potential formulation. This formulation involves the assumption of a one-dimensional magnetic field. The assumption is justified in its original frame: rather long arcs (about 10 mm), and when the electrode tip is excluded from the computational domain. However, arcs applied to welding are shorter, and the electrode geometry is important to take into account. The present work questions the assumption of a one-dimensional magnetic field for simulating short welding arcs. We have compared three different approaches for modeling the magnetic field: three-dimensional, two-dimensional axi-symmetric, and the electric potential formulation. These models have been applied to water cooled anode Gas Tungsten Arc Welding (GTAW) test cases with truncated conical electrode tip (tip radius of 0.5 and 0.2 mm) and various arc lengths (2, 3 and 5 mm). For the axi-symmetric cases studied in the present work, the three- and two-dimensional models give exactly the same results. The one-dimensional simplification of the magnetic field turns out to have a significant unfavorable effect on the simulation results. For axi-symmetric welding applications, it is argued that the two-dimensional axi-symmetric formulation should be used. Copyright © 2013 ASM International® All rights reserved.