3 ms·
Well the system prompt is still the same for both models, right? Kinda points to people at OpenAI using o1/o3/o4 almost exclusively. That's why nobody noticed
by practice9 1y ago
Well the system prompt is still the same for both models, right?
Kinda points to people at OpenAI using o1/o3/o4 almost exclusively.
That's why nobody noticed how cringe 4o has become
- astrange 1y agoThey have different uses. The reasoning models aren't good at multi-turn conversations. "GPT-4.5" is the best at conversations IMO, but it's slow. It's a lot lazier than o4 though; it likes giving brief overview answers when you want specifics.
- whimsicalism 1y agopeople at OAI definitely use AVM which is 4o-based, at least