benchmark tasks are a meme except for one: how much I am using your model through my custom nvim framework
4,55K