In what way? This was a geometric mean of the improvements from a small test corpus. In production, where it only parses longer SQL that didn't hit the parser cache, the mean parse time went down by 454x, across millions of parses.