We present four model series of the CODEX dynamical opacity-sampling models of Mira variables with solar abundances, designed to have parameters similar to o Cet, R Leo and R Cas. We demonstrate that the CODEX models provide a clear physical basis for the molecular shell scenario used to explain interferometric observations of Mira variables. We show that these models generally provide a good match to photometry and interferometry at wavelengths between the near-infrared and the radio, and make the model outputs publicly available. These models also demonstrate that, in order to match visible and infrared observations, the Fe-poor silicate grains that form within 3 continuum radii must have small grain radii and therefore cannot drive the winds from O-rich Mira variables.Details of the parameter choices for the model series are given in Section 2, along with the basic comparison of the model light curves to observations. Model predictions for spectra, in particular the effect of extension on spectra, are tested in Section 3, and the models are compared to observations of o Cet including infrared and radio interferometry in Section 4. In Section 6, we compare the model structures to previously published ad hoc molecular shell papers, and in Section 7 we discuss the mass-loss rates of the models and the driving mechanisms. In Section 5 we discuss the effect of input parameters on the models, and the possibility for better calibrating the input parameters so that, e.g., the mass of individual Miras could be inferred from models. Finally, in Section 8 we conclude and discuss plans for future work.A detailed description of the model construction is given in Paper I. Briefly, the models consist of self-excited grey models that determine the atmospheric pressure stratification and luminosity. The temperature profile is then re-iterated using an opacity-sampling code with 4300 wavelength points, assuming radiative and local thermodynamic equilibrium. Dust formation follows Ireland & Scholz (2006), except that we drop the Rayleigh approximation, instead replacing it by a smooth fit to the Mie approximation of spherical grains, weighting the Rayleigh scattering bywhere a is the grain radius, λ the wavelength of radiation, σ the scattering cross-section and σ R the scattering cross-section in the Rayleigh approximation. This cross-section σ M is the total crosssection weighted by (1 − μ), where μ is the impact parameter. This weighting ensures that the radiative acceleration on dust is correct in the presence of forward-scattering. The choice of free parameters was only briefly discussed in Paper I in the context of the o54 model series, based on the parameters of the prototype Mira variable o Cet. The input parameters for all four model series presented here are given in Table 1, and the reasons for their choice are given in Sections 2.1-2.4.The behaviour over 10 4 d of each non-linear pulsation model series is shown in Figs 1-4. As each model series runs for many cycles, we chose only a few typical cycles for a detailed e...
