Claude 4.6 Opus is here and it writes code that makes senior engineers nervous
Anthropic drops Claude 4.6 Opus with a 1M native context window, dramatically improved agentic capabilities, and what they're calling 'deep reasoning chains.' Early benchmarks show it outperforming GPT-5 on SWE-bench by 14 points. The model can hold multi-file project context and produce production-grade code with minimal scaffolding.