Claude新模型4.6来了，更多饭碗没了：华尔街财务、编译器、安全白帽、PPT…通通失守

Core Insights - Anthropic has launched its new model, Claude Opus 4.6, which has significantly impacted the market, causing declines in major financial data service providers and indices [1][3][4] - The model demonstrates enhanced capabilities across various domains, including financial analysis, research, and office tasks, surpassing competitors like OpenAI's GPT-5.2 in performance metrics [6][8][29] Model Performance - Claude Opus 4.6 outperforms OpenAI GPT-5.2 by 144 Elo points in the GDPval-AA evaluation, achieving higher scores in approximately 70% of cases [8][29] - In programming tasks, Opus 4.6 achieved the highest score in the Terminal-Bench 2.0 assessment, showcasing its superior task planning and debugging capabilities [11][24] - The model supports multi-language coding and can efficiently handle large codebases, reducing migration time significantly [26] New Features - Opus 4.6 introduces Adaptive Thinking, allowing the model to autonomously determine when to engage in deep reasoning, along with a context compression feature for managing lengthy dialogues [20][21] - The model maintains its pricing structure, charging $5 for every million tokens input/output, with additional fees for exceeding 200k tokens in context [34][35] Security and Collaboration - The model has demonstrated exceptional capabilities in cybersecurity, identifying over 500 previously unknown high-risk zero-day vulnerabilities during testing [41][42] - Opus 4.6 features Agent Teams, enabling multiple instances of Claude to collaborate on tasks without human supervision, exemplified by a project that produced a C compiler in two weeks [36][40] Market Impact - The launch of Opus 4.6 has raised concerns among investors regarding the potential disruption of traditional industries by AI technologies, leading to significant market volatility [1][4][3] - The model's advancements in reasoning and understanding complex domains like law and finance position it as a leading tool in the AI landscape [29][30]