The scores look similar to how it feels to use Argon: great at world knowledge, not so good at coding compared to Opus.

It's hard to understand the phased rollout of a model that isn't particularly sota, unlike Mythos a few months ago.