Core Viewpoint - OpenAI has launched GPT-4.1 for free, while Anthropic is expected to release new models, Claude Sonnet and Claude Opus, focusing on "Extreme reasoning" capabilities [1][3]. Group 1: New Features of Claude Models - The new "Extreme reasoning" feature establishes a dynamic loop between reasoning and tool usage, allowing for smarter problem handling [2]. - The model pauses and reevaluates problems when faced with difficulties, adjusting its strategy as needed [7]. - It can automatically adjust its direction if it encounters challenges or provides inaccurate answers, mimicking human thought processes [8]. Group 2: Code Generation Capabilities - For code generation tasks, the model tests the generated code and corrects errors instead of merely outputting results [9]. - The architecture of the new model is designed to adapt to various tasks and scenarios, reducing reliance on human supervision [10]. Group 3: Human-like Reasoning - The model can engage in deep reflection based on context rather than just statistical language generation [11]. - This collaborative reasoning approach brings the new model closer to human-like thinking, allowing it to reason rather than function solely as a "calculator" [12]. Group 4: Community Reactions and Testing - Some users express skepticism about the claims, suggesting potential hype, while others defend the credibility of the source, The Information [13][14]. - There are reports of a model called Claude Neptune being tested, which is suspected to be Claude 3.8 with a maximum token count of 128k [17].