The spectral absorption models of phytoplankton are established based on the in-situ measurements and the accuracy of the models is assessed with the comparisons among the models. A power function is used to study the phytoplankton absorption model with wavelength-dependent coefficients determined by the least squares method. The accuracy of the power model is assessed by three kinds of spectral error indices, which are root mean square error (RMSE), the spectral-average-relative error, the spectral average log-transformed relative error and with the average values of 0.07 m-1, -12.5% and -4.5%, respectively. The accuracy of the models is influenced by many factors such as the equation forms, the parameter values of the models, the phytoplankton species, pigments composition and particle size structure. The performances of the models are obviously different, even the same form of the model with different parameters. Two phytoplankton absorption models of Sathyendranath are validated and other three forms of phytoplankton absorption model are also developed to check whether the models are sensitive to chlorophyll a concentration and to the forms of the equations. The results show that it is the parameters of the models that dominate the performances of the models instead of the form of the equations. It is also indicated that the absorption values estimated by the models fall within the ranges of the measured data and the models are difficult to be validated without information of the phytoplankton species.