That claim keeps contradicted hard by other parties, who say Mythos beats 5.5 resoundingly on both autonomous search and discovery and creation of complex exploit chains.

There might be a harness difference, but also, this CTF-type benchmark might not capture the capability difference fully.