alt Hacker NewsI'm guessing their newer models are taking way more compute than they can afford to give away. The biggest challenge of AI will eventually be, how to bring down how much compute a powerful model takes. I hope Claude puts more emphasis into making Haiku and Sonnet better, when I use them via JetBrains AI it feels like only Opus is good enough, for whatever odd reason.
I get the same. Work has shifted to being agentic first - and whenever I use anything other than Claude Opus it seems that the model easily gets lost spinning its wheels on even the simplest query - especially with some of our more complex codebases, whereas Opus manages to not only reason adequately about the codebase, but also can produce decent quality code/tests in fairly short order.
Oddly though, when using at home I'm using Sonnet via the standard chat interface and that, whilst it will produce substandard code in its output is still reasonably capable - even in more niche tasks. Granted though that my personal projects are far simpler than the codebase I handle at work.