The current steady-state fuel consumption models are not optimal due to their large estimation error, complex structure, excess model coefficients, and relatively low predicting accuracy of fuel consumption. Based on experimental data of a passenger car’s fuel consumption, this study analyzes advantages and disadvantages of existing models, proposes a unified expression of the transient variable polynomial fuel-consumption model, introduces the Bayesian information criterion (BIC) for model selection, and establishes an instantaneous fuel-consumption model with simple structure and high accuracy. The simulation results show that, compared with the BIT-TFCM models, the root mean square error (RMSE) and the mean absolute percentage error (MAPE) of the proposed model are 6.27% and 6.43% smaller, respectively, on average. The proposed model has the advantages of simple structure, high precision, less influence from working conditions, stable performance, and easy prediction of the fuel consumption of future working conditions, hence providing a foundation for economic path planning.