AI startup Anthropic has introduced Claude Opus 4.5, hailed as the “world’s best model for coding, agents, and computer use.”
“Opus 4.5 is a step forward in AI systems’ capabilities and a harbinger of more significant changes in how work is done,” the company stated.
Comparison of Opus 4.5 with competitors in app creation. Source: Anthropic.The firm noted that the neural network “handles ambiguity and analyzes trade-offs independently.”
“When faced with a complex multi-system error, it finds a solution on its own. Overall, Opus 4.5 just ‘gets it,'” the developers claimed.
The company shared a video showing the LLM solving a puzzle using real-world action capabilities. The LLM “significantly outperforms” other solutions in everyday tasks such as deep research and working with slides and spreadsheets. Opus 4.5’s capabilities in programming across various languages. Source: Anthropic.Anthropic also emphasized Opus 4.5’s high safety standards. Analysis of “alarming behavior” — a wide range of inappropriate actions such as undertaking undesirable acts on its own initiative. Source: Anthropic.
“Our clients often use Claude for critical tasks. They want assurance that in the event of malicious attacks by hackers and cybercriminals, the AI possesses the necessary skills and ‘street smarts’ to avoid issues,” the company blog stated.
Opus 4.5 has received enhanced memory for long-context operations. These changes have enabled the implementation of an infinite chat feature — the dialogue can continue uninterrupted even when the model reaches the context window limit. It will compress data without notifying the user. Opus 4.5 is available across all company applications, on cloud platforms, and via API. The price for developers is $5/25 per million input and output tokens.
New Tools
Additionally, Anthropic has released three new features that allow AI agents to work more effectively:
Tool Search Tool — gives Claude the ability to use search tools without occupying the context window;
Programmatic Tool Calling — invokes necessary solutions in the code execution environment;
Tool Use Examples — provides a universal standard for demonstrating effective use of necessary tools.
“These features open new possibilities for what you can create with Claude,” the developers noted.
