Poor Experience with o3. Pleas share your experiences

I usually use it to discuss new ideas in scientific fields, and I really loved o1 for those discussions.(o1 pro is too slow… for this purpose) However, that service has been discontinued, so I tried using o3. Unfortunately, my experience with o3 has been the worst.

First, instead of answering my question adequately, it simply summarizes the answer and steers me to the next question. Regardless of the point I raise, o3 guides the conversation toward whatever it wants to talk about. If, before o3 moves on to its preferred topic, I ask again about the part I still find unclear, two problems arise:
(1) It tries to steer me toward the detailed answer it wants to give, leading to miscommunication.
(2) It treats my follow‑up as an entirely new question, gives another brief reply, and then pulls the conversation back to what o3 wants to discuss.

The issue with (2) is that the first reply is over‑summarized, leaving out key details or containing numerous errors. Perhaps because of this, I encounter hallucinations in o3 far more often than in o1, where I barely noticed any. Although people say o3’s accuracy has improved, in practice its usefulness is much lower than o1’s. Overall, it feels like talking to a self‑important undergraduate who answers however they like.

o3 is a huge step back. o3 gives quotes in answers and provides links. I started checking them and every single one contained fabricated information. And the model insists over and over that the answer exits in the source.

The hallucintion rate is 100%. This is not a joke. I can’t rely on anything. I can’t find one answer that has no hallucination.

The problem with o1 pro, busides being slow, is that you can’t easily check because it doesn’t give you sources. But o3 is definitely worse.

I really miss o1.
o3 feels less like a reasoning model and more like a hallucinating fiction writer.
It repeatedly loses track of context and invents details that were never provided, even worse than 4o.

1 Like