Gpt4o and o1 are absolutely not better at coding than Claude. They're noticeably worse at following complex, detailed prompts (e.g. a detailed prompt for it to create an application), and often forget details during the conversation (e.g. will revert to an old version of some code it already refactored). o1 is better than Claude for Leetcode hard style coding problems, but the majority of coding work isn't about that, it's about correctly implementing a spec. Plus even o1 will still often fill code with "implemention here" comments, in spite of being explicitly asked to provide a full working implementation.