Comment by 6thbit

Instead of benchkmarks it'd be useful to get a sort of Radar chart across measured capabilities like computer use, communication, coding, etc.

We're probably headed to further specialized agents where a single ordering or even model size won't mean a lot.