What is Claude Opus 4.6?

Claude Opus 4.6 is Anthropic's most advanced model, designed for deep reasoning, sustained agentic tasks, and large-scale codebase work. It features a 1M token context window in beta, adaptive thinking capabilities, and improved planning skills. The model achieves state-of-the-art performance on benchmarks like Terminal-Bench 2.0, Humanity's Last Exam, and BrowseComp, while also excelling at real-world knowledge work tasks in finance, legal, and other professional domains.

Who it's for

• Developers working on large codebases who need reliable code review, debugging, and the ability to sustain long agentic sessions without hand-holding
• Knowledge workers who regularly run financial analyses, conduct research, or create and edit documents, spreadsheets, and presentations
• Teams building autonomous workflows who want a model that can multitask independently within Cowork or assemble agent teams in Claude Code

Key features

1M token context window

Opus 4.6 introduces a massive 1M token context window in beta, allowing the model to process entire codebases, lengthy documents, or extended conversations in a single session. This makes it possible to work with large-scale projects without constantly reloading context.

Adaptive thinking and effort controls

The model can now pick up on contextual clues to determine how much extended thinking a task requires. Developers can also dial effort up or down using the /effort parameter, giving fine-grained control over the balance between intelligence, speed, and cost.

Agent teams and compaction

In Claude Code, you can assemble agent teams that collaborate on tasks together. On the API, compaction lets Claude summarize its own context to perform longer-running tasks without hitting limits, making sustained autonomous work more practical.

Improved real-world work capabilities

Substantial upgrades to Claude in Excel and a new research preview for Claude in PowerPoint make Opus 4.6 significantly more capable for everyday professional tasks like creating spreadsheets, building presentations, and running analyses.

What stands out

Claude Opus 4.6 is the strongest model Anthropic has shipped. It takes complicated requests and actually follows through, breaking them into concrete steps, executing, and producing polished work.

This isn't just about raw benchmark scores—Opus 4.6 outperforms the next-best model on economically valuable knowledge work by around 144 Elo points. Early access partners report that the model works autonomously without hand-holding, succeeds where previous models failed, and fundamentally changes how teams approach complex projects. Combined with a safety profile that matches or exceeds any other frontier model, it delivers both capability and reliability.

Worth checking out if…

You need a model that can handle deep reasoning tasks, work autonomously across long sessions, and manage large codebases or complex documents without constant supervision. If you're building agentic workflows, running professional analyses, or want a model that plans carefully and catches its own mistakes, Claude Opus 4.6 is worth evaluating.