Just tested the complex text processing capabilities of gpt-5 and gpt-5-mini (lots of tokens, complex rules, new knowledge). I asked their big brother gpt-5-pro to evaluate the results of two gpt-5-mini and one gpt-5 from multiple dimensions, and the result is gpt-5 came in last... I really don’t know what to do anymore...