First, I doubt that extrapolating from 250 kHz to 10 MHz is appropriate in any event.
However it could be, if you identified the model as for example a state space model (use ssest), then use compare to check it, and continue refining the model as necessary. You could then use findstates with lsim and an appropriate time vector, then take the Fourier transform (fft) of the result. This would be appropriate if your model has non-zero initial conditions. If it has zero initial conditions, simply use bode (see Bode Plot of Identified Model) and specify the frequency range.