Hey everyone,
I’ve noticed something strange with the newer OpenAI models—specifically o3, o4-mini, and o4-high. When I ask them to give me the full code after making changes, they often just return the diff or the changes only, not the complete modified code like I asked.
This wasn’t the case with older models like o1, o3-mini, or o3-high. Those used to follow instructions exactly and always gave me the full code when I asked for it.
Is anyone else running into this issue?
Could be a shift in instruction-following behavior or just something weird going on.
Would love to hear if others are seeing the same thing—or if anyone has a workaround.
Thanks!
2 Likes
Yes, my friend, it’s a recurring issue. You can search for other topics like:
“O3 and o4-mini are lazy. Not suitable for coding at all”
“ChatGPT (o3, o4-mini-high and even o1-pro) sucks now”
Unfortunately, it seems the company nerfed the products or restricted some accounts. I’m going to buy a new, fresh account to see if the models are nerfed there too.
2 Likes
Yes, it’s terrible. It is impossible to write long texts, they have greatly reduced the power. Previously, through o1 or even o3 mini hi, I could write texts of 2000 words or more, new models refuse to do this. Apparently we’ll have to look for a replacement for the gpt chat
3 Likes
Completly agree the new o3 model and o4 mini if gaves large code more 500 lines it will forget more than half output is super short completly useless
2 Likes
I’m confident this will be fixed in the next update, since it doesn’t make any sense as it stands. The company will either take action or at least keep us informed. Leaving customers at a disadvantage without any warning would be terrible—but I trust ChatGPT, because it’s always improving. So I see any significant setback as a bug that needs to be corrected.