I wasn't scientific about it, unfortunately. My searches were natural language, not token-based, though.

yeah, when i saw that they have 100% coverage with 1M token, i thought this must be a placeholder image, for when the actual results come in.

Because there is no variation, nothing.