logoalt Hacker News

BobbyJotoday at 2:05 AM2 repliesview on HN

The model card for mythos shows it being an incremental improvement in all respects besides security.


Replies

hodgehog11today at 3:24 AM

This is utterly daft to say if you actually used the thing for hard problems, something that benchmarks have been known to be unable to capture. It is night and day compared to Opus and every other model out there. It was nice while it lasted.

show 1 reply
bonsai_spooltoday at 2:08 AM

Ah yes, the model card that shows an over 10% improvement in agentic coding among other things!

https://www.anthropic.com/news/claude-fable-5-mythos-5