In mid-April, OpenAI launched the GPT-4.1 model, which it claims excels in following instructions. However, independent evaluations suggest that GPT-4.1 shows less alignment and reliability…