I am a heavy user of both. Where Astra, Sol 5.6 and 6.1 still significantly and materially beat Anthropic models is code review and bug detection. In the real world Opus 5.5 and Fable 5.1 miss many bugs on large code bases.

→ View original post on X — @deeplearn007