These results don’t just contradict more serious benchmarks, they are wrong on an entirely different axis. This is a saturated benchmark. Haiku gets 96%. The results here are “not even wrong” and this being #1 on HN right now is a massive smell of either bots or massive ignorance or both.