Anthropic Launches Claude Opus 4.5, Its Most Capable AI Model Yet

Anthropic has unveiled Claude Opus 4.5, the new flagship model in its Claude 4.5 series and the company’s most advanced and safest large language model to date.

The release follows the recent launch of Sonnet 4.5 and Haiku 4.5, which serve as midrange and entry-level counterparts.

Opus 4.5 is designed to excel at powering AI agents that use external tools to automate work. According to Anthropic, agents built on the model can iteratively improve when they encounter difficult tasks, reaching “peak performance” after just four iterations, compared to about ten for competing systems.

The model also offers stronger support for long-running agentic workflows, making it well-suited for tasks such as multi-hour application rewriting.

Developers can combine models in the 4.5 lineup for cost-efficient automation—using Opus 4.5 as a lead agent and Haiku 4.5 for simpler sub-tasks. Anthropic also reports notable improvements in programming performance: Opus 4.5 requires less human guidance and is better at resolving ambiguity, enabling it to troubleshoot issues across multiple systems without explicit instructions.

The flagship model arrives alongside expanded integrations. Opus 4.5 now works with Microsoft Excel, where users with Max, Team, and Enterprise plans can access Claude via a sidebar, create pivot tables, upload files, and generate charts.

The previously previewed Chrome extension, which allows Claude to take actions inside web apps, is now available to all Max subscribers.

Additional updates include higher usage limits for Max and Team plans, the addition of Claude Code to the desktop client, and improved chat summarization. Opus 4.5 is available through Claude Chat, Claude Code, and APIs, priced at $5 per million input tokens and $25 per million output tokens.