4 Likes
I wonder if that explains all the slowness?!
2 Likes
think you’ll also be happy with o3-pro pricing for the performance
Definitely looking forward to the release of that one, especially when it is possible to fire off a few requests without breaking the bank.
Let’s hope for the release very soon!
2 Likes
Still not giving that kid my ID.
Any word on o3-mini price drop?
Update:
I just wish there were more information about when. Should I have a cup of coffee and stay up, or not?
1 Like
Slightly skeptical that suddenly o3 is reduced by 80% and then o3-pro is announced at the same time. Is this not just … Basically o3-mini vs o3?
Is this the same concept as naming regular chips “Family Size”, and then reducing the regular chip size?
1 Like
Any news on an upcoming full o4?
Right now o3-mini and o4-mini are priced the same, if a full o4 comes out at the current o3 prices that would be game changing.
1 Like
Actually, I’m expecting GPT-5 to be released fairly soon, and I have to admit that from that point on, I stopped thinking about o4 
.
Also, I was expecting GPT-5, but I need a thermos of coffee.
BTW they posted release note:
OpenAI o3-pro, available now for Pro users in ChatGPT and in our API.
https://help.openai.com/en/articles/6825453-chatgpt-release-notes#h_6c98aab643
2 Likes
So I had to look back at the o3 announcement page.
Notice that o3 scored 2706 & o4-mini scored 2719
Now o3 scores 2517 and o3-pro is 2748?
Reduced to 2301 from 4/4 reliability benchmark
At this point, o4-mini is most likely going to be better & cheaper than o3 as it so far looks to be a distilled model. Could be why they only compared it to o1-pro and the now reduced o3?
1 Like
Probably just one more “A model version we’ll never release to the public” benchmark.
I’m actually sad from this release. I imagine this new o3 will replace the current one for Plus users.
Dark patterns FTW
There is a comment regarding the updated code forces eval. Apparently the numbers can’t be compared 1:1.
- The Codeforces evals for o3 and o3-pro were run using an updated set of Codeforces questions with more difficult tasks, as the previous version (used for o1-pro) was close to saturation.
1 Like
Still expensive but much more reasonable, o1-pro was totally out of reach for me.
I honestly picked the first one I saw.
This pattern is consistent throughout the shown benchmarks:
o3 Announcement page
New o3-pro page
The same pattern: o4-mini outperforms or is competitive with o3 and is almost 1/2 the price
- The Codeforces evals for o3 and o3-pro were run using an updated set of Codeforces questions with more difficult tasks, as the previous version (used for o1-pro) was close to saturation.
Where did you find this?
Ah, thank you.
It would’ve been nice to have o4-mini on the graph. At best, it’s highly ambiguous on which o3 model they used.
1 Like
