OpenAI发布其最强模型GPT-5.4，直指Anthropic核心市场

Core Insights - OpenAI has launched its new AI model GPT-5.4, which is claimed to be the most capable AI system for professional scenarios to date, integrating advanced reasoning, programming, and autonomous operation capabilities [1] - The model combines various capabilities previously spread across different models, including programming abilities from GPT-5.3-Codex and enhanced reasoning skills, allowing it to operate autonomously across desktop, browser, and software applications [1] - The release of GPT-5.4 is expected to intensify competition in the enterprise AI market, previously dominated by Anthropic [1][2] Model Features - GPT-5.4 introduces "out-of-the-box" agent capabilities, enabling it to autonomously operate computers and software, search for and call external tools, and handle complex multi-step tasks without requiring developers to build underlying infrastructure [4] - Compared to GPT-5.2, the new model has significantly reduced the probability of generating "hallucinations," with a 33% decrease in the likelihood of individual statements being incorrect and an 18% reduction in overall error probability [4][5] - The model has shown improvements in token usage efficiency, allowing it to solve problems with fewer tokens, which may offset the slight increase in token pricing for some users [5] Competitive Landscape - OpenAI's new model release has raised concerns among investors regarding the potential disruption of traditional financial data providers by AI technologies, leading to previous declines in SaaS stock prices [3] - OpenAI has also introduced ChatGPT versions for Excel and Google Sheets, integrating AI directly into spreadsheets for building and analyzing complex financial models, further intensifying competition with Anthropic [2] - The market is witnessing a growing preference for AI systems capable of managing long workflows with minimal human intervention, as evidenced by the rise of products like OpenClaw [5]