This is a modeling problem:
The "capacitance" is not a true capacitor, but a distributed (transmission line) effect.
We can construct a model which approximates it, but which is not physically representative.
To that end, the most accurate models I know of are by Coilcraft. This page gives an explanation of their model design, a viewer, and a way to rewrite their models for use in all SPICE simulations.
https://www.seventransistorlabs.com/Calc/Coilcraft1.htmlCoilcraft's SPICE page is linked, where you will find parameters for most of their parts. If you are looking at another brand, it may be close enough to look for a Coilcraft cross-reference.
Downside: most of the models stop abruptly after (and sometimes before) the self resonant frequency. Therefore, we should suppose that these models are not very accurate above the SRF, and the capacitance given is only a fit for that range, not a physical value.
In practice, there will be numerous dips and peaks (series and parallel resonances) above the first SRF, depending on the exact placement and number of turns in the part. These features cannot be represented with the above model.
Tim