I'd prefer it if our benchmarking groups were 100% transparent about how their testing stack works and what models they are including in the mix

→ View original post on X — @simonw