3 ms·

I am also trying to understand the shady limits of LLMs. But your example doesn't give incorrect answers in ChatGPT 4o, Sonnet 3.5 nor Deepseek V2.