They classify the models upon reproducibility and openness criteria. So a freely accessible model lacking proper documentation/reproducibility is essentially se

They classify the models upon reproducibility and openness criteria. So a freely accessible model lacking proper documentation/reproducibility is essentially seen as non-free: "toxic candy" follows all Free Software good practices, yet the model they depend on suffer from such issues. That would mean if there was e.g. a particularly sneaky input data bias, it could cause issues down the line where it would be hard to determine if it was caused by a poorly calibrated parameter or missing training examples.