Core Viewpoint - The article discusses the significant upgrade of Claude AI with the release of Claude 4, featuring two main models: Claude Opus 4 and Claude Sonnet 4, which set new standards in coding, advanced reasoning, and AI agent capabilities [1][2]. Model Features - Claude Opus 4 is highlighted for its exceptional endurance, capable of running independently for 7 hours while performing coding tasks, a substantial improvement over its predecessor, Claude 3.7 Sonnet, which could only operate for 45 minutes [5][7]. - Claude Sonnet 4, while similar to Opus 4, is optimized for reasoning and efficiency, resulting in lower operational costs [19]. - Both models are hybrid, offering two operational modes: one for quick responses and another for deeper reasoning [20]. Performance Metrics - Claude Opus 4 achieved a score of 72.5% on SWE-bench and 43.2% on Terminal-bench, outperforming all previous Sonnet models [20]. - Claude Sonnet 4 scored 72.7% on SWE-bench, demonstrating its competitive capabilities [21]. - The models excel in coding, reasoning, multimodal capabilities, and agent tasks [23]. Pricing Structure - The pricing for Claude Opus 4 is set at $15 per million tokens for input and $75 for output, while Claude Sonnet 4 is priced at $3 for input and $5 for output [24]. New Functionalities - The release includes new features such as code execution tools, MCP connectors, file APIs, and a one-hour cache prompt capability to enhance developer experience [16][29]. - The models can utilize tools during reasoning processes, improving their ability to follow instructions and maintain continuity over time [30]. Strategic Shift - Anthropic has shifted its focus from chatbots to enhancing Claude's ability to execute complex tasks, including research and programming, which has led to significant advancements in coding capabilities [47][49]. - The company has secured a $2.5 billion revolving credit facility to support its AI research and training efforts, reflecting strong market confidence [55]. Revenue Growth - Anthropic reported an annualized revenue of $2 billion in the first quarter, more than doubling from the previous quarter's $1 billion, with the number of customers spending over $100,000 annually increasing eightfold [56][57].
最强编码模型Claude 4!7小时不间断写代码,连玩24小时宝可梦,GitHub已选为Copilot底层模型
量子位·2025-05-23 00:24