Anthropic, a leading developer in artificial intelligence and AI safety, made a significant impact on May 22, 2025, by announcing the next generation of its flagship language models – Anthropic's Claude 4. This major update introduced two advanced models: Claude Opus 4, designed for tackling the most complex intellectual tasks, and Claude Sonnet 4, offering an optimal balance of performance and speed for a wide range of enterprise applications. This release marks a substantial step forward in AI capabilities, particularly in coding, logical reasoning, and autonomous tool use.
The new Anthropic's Claude 4 models demonstrate significant progress across all key areas. Developers highlight considerably improved abilities in writing and debugging code in various programming languages, as well as enhanced accuracy and depth in logical reasoning when solving complex multi-step problems. Particular attention was given to "agentic functions," meaning the models' ability not just to generate text, but to interact with external tools and systems to achieve set goals. In this context, Anthropic introduced a beta version of an innovative "expanded tool use thinking" feature. It allows the models to dynamically switch between internal information processing and active use of external resources like web search, calculators, or specialized APIs for more effective problem-solving.
Claude Opus 4, as the most powerful model in the lineup, has gained the unique capability to access memory files (with appropriate user permission). This enables it to form a deep contextual understanding based on provided documents and maintain consistent, relevant responses in lengthy dialogues or when working on extensive projects. Furthermore, both Opus 4 and Sonnet 4 models now support parallel execution of multiple tools, significantly speeding up complex tasks that require several operations. Improved memory processing and more precise instruction following further enhance the reliability and predictability of the AI's performance. Aware of the potential risks associated with the increased capabilities of its models, Anthropic has activated its Responsible Scaling Policy (RSP) and implemented stringent ASL-3 (AI Safety Level 3) security measures for Anthropic's Claude 4, reaffirming its commitment to safe AI development principles.