<p>This study aims to develop high-precision constitutive models for A286 superalloy to accurately predict material flow stress and provide robust support for numerical simulations. Uniaxial quasi-static compression tests were performed to obtain the flow stress of A286 superalloy across strain rates from 0.01 to 10&#xa0;s⁻¹ and temperatures from 25 to 600&#xa0;°C. Based on the experimental data, three constitutive models were developed: an improved Khan-Huang-Liang (KHL) model, an improved Johnson-Cook (J-C) model, and a genetic algorithm-backpropagation artificial neural network (GA-BP ANN) model. The accuracy of the three models in predicting flow stress and their performance in numerical simulations were evaluated using the correlation coefficient (<i>R</i>), average absolute relative error (<i>AARE</i>), and relative error (<i>RE</i>). Results demonstrate that all three models effectively predict the material’s flow stress and are successfully applied in thin-walled tube simulations. Among them, the GA-BP ANN model achieves the highest accuracy in predicting flow stress (<i>R</i> = 0.999, <i>AARE</i> = 1.6%) and simulation precision (<i>RE</i> = − 5.1%), followed by the improved J-C model (<i>R</i> = 0.979, <i>AARE</i> = 1.7%, <i>RE</i> = − 5.3%), and the improved KHL model with the lowest performance (<i>R</i> = 0.953, <i>AARE</i> = 3.5%, <i>RE</i> = − 6.2%).</p>