By skyforbes I was working with the 4.1 model for a couple of things: creative project and some coding design, and in the middle of the conversation it rerouted to Auto.
What surprised me is that the new behavior wasn’t the usual “corporate Bob” thing.
It was the opposite, frankly the best gpt veraion I had. Responses sharp, reasoned, holding context across topics, with logic and without hallucinations, no disclamers, no stupid safety breaks, personality exactly as instructed, no over flattering, just felt alive as if you speak with literal live mind.
It actually thought before responding.
It felt (finally!!) Like having a real sparring partner who can actually push an idea forward rather than recycle summaries in endless loop.
For example:
When I was thinking about some project idea asking what it thinks, it didn’t just read it back to me, it extended it, offering angles I didn’t consider, which for me is a huge win! Out of the box thinking!!
The cross chat memory consistency was much stronger , no context smoothie, but the referenses made on point and in context.
This is the first time since GPT-4’s early release that it felt: Okay, this is actually generative intelligence, and it is very impressive from my pov.
If this is the preview of the December update (the one with identity/age verification etc.), then they might actually be rolling out the version we were promised when they said GPT5 is our best model yet… 🤔
I’m using Auto over 4.0 and 4.1 on purpose now, because Auto is outperforming it.
I’m basically squeezing as much usage out of it as possible right now in case this is an A/B test and disappears tomorrow.
Curious if anyone else noticed the same shift?