Design standards for composite structures present safety verifications for the most relevant failure modes that can occur in civil engineering structures. However, in some cases, such as the local buckling of pultruded fibre-reinforced polymer (FRP) profiles under flexure, the experimental data available in the literature is scarce, not allowing to conduct comprehensive reliability analyses of such resistance formulae. In this context, the aim of the investigation presented in this paper is to provide additional experimental data about this phenomenon and to compare it with predictions obtained from (i) the resistance models included in different design codes, and (ii) computational models. To that end, pultruded rectangular hollow section (RHS) glass-FRP (GFRP) beams were tested in a 4-point bending configuration that promoted the triggering of local buckling phenomena. Experimental results were compared with predictions obtained from local buckling formulae included in the American, Chinese and European design codes, allowing to assess the bias between experimental and analytical results. Additionally, non-linear finite element (FE) models were developed to predict (i) critical and (ii) failure loads, while critical loads were also estimated with a finite strip (FS) model. These simulations allowed to assess not only the accuracy of the computational models in predicting the critical loads, but also the post-critical strength reserve of pultruded GFRP beams with RHS section (FE models only).