Current state of LocalLLaMa
Posted by IngenuityNo1411@reddit | LocalLLaMA | View on Reddit | 6 comments
[yeah, Claude 3.5 sonnet and GPT-4...]()
Seems like bots want to involve despite they thought we're still living at somewhere 2024
Ornery-Peanut-1737@reddit
fr the speed of innovation is kind of overwhelming, honestly. every time i get my infrastructure settled and a nice workflow built out, a new model drops that renders my old setup basically obsolete. it is fun, but also kind of exhausting trying to stay on the bleeding edge. my strategy has just been to stick to a stable backend for a few months and only update when i actually need a feature that the new models offer. it saves a lot of headache compared to chasing every single hype release.
IngenuityNo1411@reddit (OP)
Sure you can't always stay on bleeding edge, but nobody here would talk about claude 3.5 sonnet or GPT-4 and compare them with recent open weight models. we all knew there are claude 4.5 sonnet and GPT-5.4, but LLM-powered bots don't.
IngenuityNo1411@reddit (OP)
You get it wrong, what I mean is bot posts are flooding this reddit and one obvious sign is they tend to refer outdated models around one year ago (which is the typical knowledge cutoff date of current LLMs, and these posts are generated by them)
DinoAmino@reddit
You're doing it wrong then by trying to stay on the bleeding edge.You could never be productive switching out models all the time. New models being released doesn't't mean you have to start over. One of the worst misconceptions on this sub is that old models are useless - unless you rely solely on internal knowledge being "current", truth is they are already 6 months of date after release.
LocalLLaMa_reader@reddit
It's always the same complaint. Just downvote or report it, these posts will be wiped away soon.
BothYou243@reddit
seeing one of the best sources of info: reddit , driven by bots hurts