Anthropic’s Claude Opus 4.1 Improves Refactoring and Safety, Scores 74.5% SWE-bench Verified - ai_code | Inventtive | Inventtive