Problem with Muse Spark 1.1
I was checking benchmarks of muse spark and first it looked impressive that it scores 80 on terminal bench 2.1, but then i saw vals ai's post: in which they are showing that muse spark scores 69.29, which again proved my hypothesis true that benchmakrs are really inconsitent and does not tell whole story.
how to trust these benchmarks?
what would you say? or i am missing smth? oh, facebook has a reputation with benchmarks. who remember massive llama 4 failure?