AI hallucinations are getting worse – and they're here to stay

6 OutOfHere 1 5/10/2025, 2:36:11 AM newscientist.com ↗

Comments (1)

OutOfHere · 3m ago
In my experience, this is an issue with the newer reasoning models only, e.g. o3, o4-mini. It is not an issue with gpt-4.5.

o3 loves to hallucinatine a couple of assertions toward the end of a lengthy response.