Hi everyone, I’m running into a weird issue and hoping someone can help or explain what’s going on @OpenAI_Support
Whenever I try to use the newer reasoning models (like o4-mini and o3) on the web version of chatgpt, the way they process and display their “thinking” seems off. Instead of using the new reasoning interface (where it shows a structured “thinking” plan in a neat little box), it behaves like the older o-series models and when asked what model capabilities it has it says it’s based on gpt 4 mini version?. It starts saying “reasoning:” and then just reasons out loud step by step in the old formatting style.
It doesn’t seem to follow instructions as clearly either, almost like it’s using the logic of older models instead of taking advantage of the updated reasoning capabilities and tools.
I’ve only encountered this issue on the web. On the mobile app, it works exactly as expected of the new models, with the proper layout and reasoning capabilities. I haven’t tried on the desktop app, but web seems to be the issue. I have a Plus account.
Does anyone know what’s happening? Is this a bug, or a settings issue? I tried disabling and enabling 2fa, didn’t work. I’ve tried searching around, but haven’t found anything relating directly to this issue.