o4-mini (the reasoning model) is interesting to me, it’s like if you took GPT-4 and stripped away all of those pleasantries, even more so than with GPT-5, it will give you the facts straight up, and it’s pretty damn precise. I threw some molecular biology problems at it and some other mini models, and while those all failed, o4-mini didn’t really make any mistakes.
o4-mini (the reasoning model) is interesting to me, it’s like if you took GPT-4 and stripped away all of those pleasantries, even more so than with GPT-5, it will give you the facts straight up, and it’s pretty damn precise. I threw some molecular biology problems at it and some other mini models, and while those all failed, o4-mini didn’t really make any mistakes.