A key development in metal AM technology is the emergence of LPBF machines equipped with up to 8 lasers to melt the metal powder simultaneously. Multi-laser machines are aimed to speed up the printing process and increase the limited part sizes in LPBF. However, temperature transients in the part can be significantly different compared to a single laser scanning process depending strongly on the individual scanning of each laser and the mean separation. Therefore, we envision computational prediction of the temperature transients of a part during the build, which is paramount to gaining insight into how to design process conditions in multiple laser AM processes. For that purpose, computationally efficient thermal process models that account for multiple lasers and quantify the effects of individual laser speed, power and scanning strategy on thermal history are presented in this paper. We utilise a computationally efficient multi-laser thermal process model [1] for LPBF, which describes the moving laser spots with sets of point heat sources along the laser scanning vectors. We then utilise the closed-form analytical solution for a point heat source in a semi-infinite medium to capture the steep temperature gradients analytically in the vicinity of the laser spot transiently. Boundary conditions of the problem are then enforced utilising the superposition principle, i.e. with the combination of image point heat sources situated outside the domain and a complementary smooth correction field. The former are also described analytically, while the latter is solved numerically. Since the steep temperature gradients near the laser spot are accounted for with an analytical description, a coarse spatial discretisation for the complementary numerical correction field becomes suitable. Consequently, computational costs of the thermal process simulations are significantly reduced, enabling simulations of multiple layers for parts with dimensions in the cm length scale. Moreover, the simulation of multiple laser scales favourably with this approach. However, so far, image fields can be applied to bodies with convex surfaces, limiting the application of the above-described approach to simple geometric layouts unless a mesh refinement is performed near the part's surfaces [2]. Therefore we will present a new formulation for the image fields that extends this approach by determining a correct image source position with the local curvature of the boundary.