In order to investigate the behaviour of a steel structure and provide effective design, most of the time analyses procedures which can take account the non-linear material behaviour are needed. These non-linear analysis procedures which are widely applied with different levels of accuracy, adopt two major approaches which can be classified as lumped plasticity and distributed plasticity models. This paper aims to analyze the accuracy of two different plasticity approaches such as lumped and distributed plasticity models on providing the non-linear material behaviour of steel elements under different loading protocols by comparing numerical and experimental results. In the first part, monotonic and cyclic quasi static tests and pseudo dynamic tests under instantaneous constant load and seismic load are conducted by using the small scale Load and Boundary Condition Box. Small scale steel specimens are used for the experiments. Results of the experiments are comprehensively reported. In the second part of the study, several numerical models are generated for each test. Monotonic and cyclic loading pushover analyses are carried out with both lumped plasticity models and distributed plasticity models. In these numerical studies, different types of material models are considered and numerical analysis results are validated against the experimental results. Accuracy of lumped plasticity and distributed plasticity models are assessed and a comprehensive comparison of different analysis and material models is reported. It is generally observed that using the fiber based method with bilinear kinematic hardening material model provides more accurate element responses for hot rolled steels.