Very limited research studies have been conducted to examine bond of glass fibre reinforced polymer (GFRP) bars with high concrete strength. The current research project aims to compare between bond measured from a pull-out test and a hinged beam test for GFRP bars embedded in high strength concrete. Different parameters influencing bond such as GFRP bar diameter, embedment length and surface configuration were investigated in both test methods, while the bar position, i.e. top or bottom, was only studied in hinged beams. Seventy-two pull-out cubes, eight pull-out prisms and twenty-four hinged beams reinforced with GFRP bars were constructed and tested to failure. Twelve pull-out cubes and four hinged beams reinforced with steel bars were also tested for comparison purposes. The results showed that bond stress – slip curves obtained from various testing methods were similar, consisting of high initial stiffness, followed by nonlinear ascending and softening branches. In addition, it was found that the experimental bond strength obtained from hinged beams was higher than both bond strengths measured by the pull-out cube and pull-out prism. However, when a finite element analysis was conducted for hinged beams, it was shown that the tensile force in the reinforcing bar estimated by equilibrium conditions is overestimated as the large deformation of hinged beams at failure was not considered. Therefore, if the tensile force obtained from the finite element analysis is used to calculate the bond strength, it would be similar to that obtained from pull-out cube and prism. Moreover, it was found that the distribution of tensile and bond stresses was nonlinear along the GFRP embedment length and bond stress at the vicinity of the free end increased with increasing the load due to redistribution of bond stresses along the embedment length. Bond strengths were compared against the prediction methods provided in ACI-440.1R, CSA-S806, CSA-S6 and JSCE 1997. In general, all design codes showed conservative results for all specimens tested and ACI predictions gave a good agreement with experimental data compared to other codes. Artificial neural network models were developed to predict bond strength of GFRP bars in concrete. These models used bar diameter, embedment length, concrete compressive strength and concrete cover as input variables. The developed ANN models showed to be able to predict bond strength of GFRP bars in concrete and, therefore, were used to conduct a parametric study. / Higher Education Institute, Government of Libya