claude-opus-4-6

Claude Opus 4.6 is positioned for demanding coding, agentic workflows, and professional knowledge work. Anthropic says it plans more carefully, sustains longer-running agent tasks, works more reliably across large codebases, and improves code review and debugging by catching its own mistakes. These capabilities make it a practical fit for complex software maintenance, autonomous task execution, and work that benefits from extended reasoning.

The model is described as the latest and most intelligent Claude Opus release, with state-of-the-art results on several evaluations. Anthropic reports the highest score among compared models on Terminal-Bench 2.0, leads frontier models on Humanity’s Last Exam, and leads on BrowseComp. It also reports a roughly 144-Elo advantage over the next-best model on GDPval-AA, alongside improvements in financial analysis, research, and creating or using business documents.