Reliable prediction of the transient temperature field is essential for controlling melt-pool geometry, cooling rate, microstructure, and ultimately part quality in directed energy deposition-laser beam (DED-LB). Conventional gaussian double ellipsoidal (GDE) heat source model, developed largely for welding and coaxial configurations, cannot capture the asymmetric energy input of an off-axis beam or the homogenising effect of the powder stream of a top-hat laser profile. This study proposes a modified double-ellipsoidal heat source that combines a top-hat energy distribution with an explicit geometric correction for off-axis laser. The model was numerically implemented in Abaqus (an FEA software) using a DFLUX subroutine and progressive element activation, and validated against single track deposition of 316L stainless steel (SS316L) on a G250 mild steel substrate (3.3 kW, 2000 mm min⁻¹, 25° laser beam inclination). The numerical predictions were validated using in-situ infrared thermography, an embedded thermocouple, and metallographic measurements of the deposited track. Compared with the conventional Gaussian double-ellipsoidal model, the proposed top-hat double-ellipsoidal formulation more accurately reproduced the experimentally observed asymmetric and elongated melt-pool morphology. In particular, the predicted transverse melt-pool width differed from the experimental measurement by only 6%, compared with a 17% deviation for the conventional model, while melt-pool depth errors remained within 7–9%. Peak surface and subsurface temperatures deviated from measurements by 8.9% and 2.4%, respectively, and liquidus–solidus cooling rates along the track agreed within 10%. These results demonstrate that the proposed model provides an efficient and physically representative approach for simulating thermal fields in off-axis DED-LB and offers a useful foundation for process optimisation, multi-track deposition analysis, and thermal–microstructural property prediction.