This thing:
formatting link
I don't know that it's actually that exaggerated, but it's behavior that matches with rolled silicon steel: low initial permeability, higher dynamic permeability. Typical figures are, say, mu ~ 700 below, oh I don't know, a kG? Average mu ~ 20k up to 10kG, then saturation in the 10-20kG range.
Or 0.1, 1 and 1-2T in non-heathen units. :^)
This is a wide difference, which is mostly attenuated by the distributed air gap. Evidently, the exact figures for this material (that is, the B-H curve of a single grain) are enough to give this response.
Note that this is not a B-H curve, but the cycle averaged slope (magnitude).
I guess it's not obvious if the cycle being averaged over, is, say, a sine wave in B, H, or some distorted waveform inbetween (say that you'd get from an LC resonant tank in the presence of this nonlinearity). The ratio depends on how much time is spent in each region, so a cuspy waveform might show a larger ratio than a smooth one; it matters, but isn't documented.
This is pretty useless for most purposes, even for yours I suppose -- check the core loss at 60Hz and 1000s gauss, it's still dissipating something! These materials (40, 26 and 52) are gnarly beasts.
That said, you may well have even smaller size if you opt for a "Hi-Flux" core, or a laminated material. If you can handle even more losses, that is.
Tim