ParentFull threadarmarr·GLM was twice as verbose running the Artificial Analysis benchmark. So it ends up being more expensiveView on HN