Anthropic has released Opus 4.5, its “safest and most advanced” model.
Anthropic has launched its “safest and most capable” model: Opus 4.5. Following the two midrange models Sonnet 4.5 and Haiku 4.5, this is the company’s new “flagship model”.
‘Faster AI Agents’
According to Anthropic, Opus 4.5 is better at powering AI agent automation tools than its competitors. When AI agents encounter a task they can’t complete on the first try, the AI model allows them to independently adjust and refine their capabilities. Performance reaches perfection after four iterations, while other LLMs require ten attempts according to the AI startup.
Agents designed for longer tasks with context receive better support than with previous models. This is particularly useful for tasks like rewriting applications, which often take several hours. Programming capabilities have also been improved. Opus 4.5 requires less human intervention and can handle unclear instructions more independently.
Integration with Excel and Chrome
Anthropic is simultaneously expanding the Claude ecosystem. Excel integration is now generally available for Max, Team, and Enterprise users, with support for pivot tables, file uploads, and charts. The Chrome extension, which can perform actions in web apps, is being rolled out more widely. The feature protects users against misleading or harmful prompts on websites.
Availability
Claude Opus 4.5 is available in Claude Chat, Claude Code, and programming interfaces for applications. Claude Code is also available in the desktop app. Developers using APIs get a new “effort” setting allowing them to set the LLM’s time and power for a specific task. Higher power leads to better results. Max and Team subscriptions receive higher usage limits, and Claude Chat can now automatically summarize previous parts of a conversation. You pay five dollars per million input tokens and 25 dollars per million output tokens.