PM6 predicts the twisted conformation of oxalic acid (HOOC-COOH) to be 2.9 kcal/mol lower in energy than the planar form, and B3LYP predicts the twisted form to be 1.4 kcal/mol higher than the planar form. There is no reason to doubt the B3LYP result, therefore PM6 has a definite error.
PM6 predicts crystalline oxalic acid to be composed of planar oxalic acid molecules, in agreement with X-Ray results.