wondeting on how do you guys think about this
I find sometimes o3-mini is worse than 4o for certain analysis
Thoughts?
3 Likes
I use o3 high and Claud for code and math , 4o for general research, analytics etc. o3 mini is good at code too . O1 is decent at math and code.
1 vote for 4o.
4o is basically faster. Simple questions do not need oX models, and who knows if the AI thinks too much and gives me an wrong answer. For some use cases, o3-mini might be better (for academic articles?), but in my use case, o3-mini (with data) did not handle well for the questions from our customers, and it was slow.
Even to get help for my codes, I still use 4o or less advanced model to get advice, they already answer well, and again they are less expensive and fast.
I mostly use 4o as it is faster, and if it is struggling with wrong answers then I go for o3-mini.
Usually o3-mini is better for solving complex coding issues.
When I need a more direct answer not related to coding, then I go with gpt-4.5 as it gives shorter answers with a better style.
I leave o1 for when I run out of o3-mini, which hasn’t happened in a while.