Oh wow that’s a huge difference in price. I assume it will come down on the next iteration.
3 Likes
Yeah, not for Production this one … yet …
4 Likes
But looks like a wonderful model overall - lower hallucination rate, some quasi CoT reasoning, nicer responses.
3 Likes
I liked the more succinct response style … but I hope that doesn’t prevent it taking on a range of characters flexibly.
2 Likes
Wow! 
It is being what I am imagining with how many times they are saying “such a big model”.
It is being even more expensive than gpt-4-32k!
Oh! Error in documentation?
I am seeing this too… so which is it being? Big difference to our banking acounts!
TBH, they could have saved us time and just posted a blog about gpt-4.5
2 Likes
hmm, I tend to cringe when software developers say they will “simplify” the experience. ugh.
(please don’t simplify, instead hide advanced options if you don’t want most users to see them - but please make them available!)
1 Like
125k context window, max 16k out.
2 Likes
that’s confusing?
what’s the breakdown?
I don’t see it. Where do you see it @francis.poisson2 ?
Maybe if you call the base gpt-4.5-preview you are being billed at the higher rate than if you call the checkpoint gpt-4.5-preview-2025-02-27?
Like with the August 4o checkpoint?
2 Likes
You have to expand the list, button at bottom of table.
(personally I don’t yet understand the distinction)
1 Like
Ok I see it now after expansion.
Looks more like a copy-paster error (lower pricing). But let’s let them know!
1 Like
that could be it … but what’s the justification for that? so they can get feedback on a specific model?
I am thinking the $150 should be being the goof!
They should be updating the pricing page fast though because that is being a legal document, being an error or not.
1 Like