I tried GPT-5.4, and most answers were really good - but a few had me concerned ...
OpenAI's new GPT-5.4 clobbers humans on pro-level work in tests - by 83% ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results