速递｜Anthropic推出Claude 4AI模型，高端模型Opus 4持续7小时输出不宕机，抢占AI编程入口

Core Viewpoint - Anthropic has launched two new AI models, Claude Opus 4 and Claude Sonnet 4, claiming they perform among the best in industry benchmarks, particularly optimized for programming tasks [1][3][4]. Model Performance and Features - Claude Opus 4 and Sonnet 4 are designed to analyze large datasets, execute long-term tasks, and perform complex actions, with Opus 4 being the more powerful model [1][4]. - Sonnet 4 is positioned as a direct replacement for Sonnet 3.7, showing improvements in programming and mathematical tasks, and better adherence to instructions [4]. - Both models are capable of parallel tool usage and can switch between reasoning and tool use to enhance answer quality [8]. Pricing and Accessibility - Opus 4 is available only to paid users, priced at $15 per million tokens for input and $75 for output, while Sonnet 4 is priced at $3 for input and $15 for output [1]. - Token is defined as the basic data unit processed by AI models, with 1 million tokens approximately equating to 750,000 words [1]. Revenue Goals and Financial Backing - Anthropic aims to achieve $12 billion in revenue by 2027, a significant increase from the projected $2.2 billion for this year [3]. - The company has secured $2.5 billion in credit financing and raised billions from Amazon and other investors to support the development of advanced models [3]. Competitive Landscape - Anthropic faces stiff competition from companies like OpenAI and Google, which are also developing powerful AI models and tools [3][4]. - Despite recent advancements, Anthropic acknowledges that its models do not outperform competitors in every benchmark [4]. Internal Testing and Security Measures - Opus 4 has undergone rigorous internal testing, revealing potential risks in enhancing capabilities related to STEM fields, prompting the implementation of stricter security measures [7]. - The models are described as "hybrid," capable of both immediate responses and deep reasoning, with a user-friendly summary of their thought processes [7]. Development and Updates - Anthropic is committed to more frequent model updates to continuously provide improvements and maintain a competitive edge [10]. - Early tests indicate that Opus 4 can operate independently for extended periods, with a notable example of optimizing open-source code for 7 hours [10].