Anthropic has released Claude Opus 4.6, dubbing it the company’s ‘most powerful mod’ yet. In a press release on Thursday, the Dario Amodei-led AI firm added that the new model update ‘plans more carefully, sustains agentic tasks for longer, can work more reliably on larger codebases and has better code review and debugging skills to catch its own mistakes’.
Anthropologist Dario Amodi CEO and co-founder during the 56th annual World Economic Forum (WEF) meeting in Davos (REUTERS)Here are 5 key points about Claude Opus 4.61. Huge 1M Token Context Window (Beta)Opus 4.6 introduces a 1 million token context window, a first for Opus-class models. This allows AI to process and reason about extremely long documents, codebases or datasets without losing track of previous information.
2. Significantly improved coding and agentic capabilitiesThe model plans more carefully, sustains longer agentic tasks, operates reliably on larger codebases, and shows stronger code review and debugging skills, according to Anthropic. It catches its own mistakes more effectively than its predecessor (Claude Opus 4.5).
3. Several benchmarksOpus 4.6 leads the industry in several assessments:
Highest Score in Terminal-Bench 2.0 (Agent Coding)
Final Exam of Top Humanities (Complex Multidisciplinary Reasoning)
OpenAI outperforms GPT-5.2 by ~144 yellow points and Cloud Opus 4.5 by 190 points in GDPval-AA (economically valuable knowledge work, finance, legal and other domains)
Best performance on BrowseComp (hard to find information online)
4. Expanded daily work abilityOutside of coding, Opus 4.6 is built for common workplace tasks:
Detailed financial analysis is ongoing
Conduct in-depth research
Creating and editing documents, spreadsheets and presentations
Autonomous multitasking within Claude CoWork, where it can handle multiple complex requests simultaneously
5. Strong security profileAccording to Anthropic’s system card, the Opus 4.6 exhibits an overall security profile as good as, or better than, any other Frontier model, with low rates of mis-connected behavior across security assessments.
Scott White, Anthropic’s head of product for cloud AI models, said the company plans to focus heavily on improving its capabilities. “Those are the areas where we’re going to lean really hard,” he said, listing cybersecurity, life sciences, healthcare, and financial services.
