It did for me in Spanish: https://imgur.com/a/p3gOOnG

Perhaps different capabilities in different languages?

It's just not deterministic, even if you were to re-run the exact same prompt. Let alone with the system generated context that involves all the "memories" of your previous discussions.

It fails in chatGPT in french too:

https://chatgpt.com/share/6992dc05-003c-8004-9f7f-c40c7fac64...

Interestingly, just typing "Think" as a response makes it get to the right conclusion:

https://chatgpt.com/share/6992dc05-003c-8004-9f7f-c40c7fac64...