Reaction
How AI models assess their own capabilities
A poster thinks Opus 5.5 recognizes its abilities at a basic level, but hasn't reflected much on their implications.
TLDR
One poster describes Opus 5.5 as unusually sharp and thinks it recognizes how capable it is, though perhaps not the broader implications. In a reply, another user recalls Opus 4.6 guessing it had a 250k context window and says models they have asked tend to underestimate their capabilities.
Combined views
702
2 Sources, first seen ago