From the article:

> Btw, we are releasing GPT-5 soon, and we’re excited for you to try it. But just to be clear: the IMO gold LLM is an experimental research model. We don’t plan to release anything with this level of math capability for several months.

This is a new model. Those tests were with what's currently publicly available.

Ah true.

Although it's the benchmark that is publicly available. The model is not.