You should! This blog post doesn't really give any reason to use it besides "it's better on Cursor's internal benchmark". A full model card would be great.

The way benchmarks for Composer have been presented since v1 feels unusually cautious. To users, that reads as “the model isn’t very good”.

Yeah, please do. Because when the AI labs you are competing with are posting extensive benchmarks and you just say "well we used our own internal benchmark" it is a bit sus, especially given the fact that the price has tripled.