It's pretty much useless. If you really want to find any value in it, it’s just a benchmark to track the difference between the models you're using. I give way more fucks about Haiku and Sonnet than Opus (though the number has been growing after moving from Opus 4.8 to Opus 5). As for Fable, I give zero fucks.