Claude Opus 4.6

Anthropic just dropped a game-changer. Claude Opus 4.6 isn't just another incremental update—it's a massive leap forward for anyone who relies on AI for serious work. Whether you're a developer wrestling with a massive codebase, a researcher sifting through hundreds of documents, or a business professional automating complex workflows, this model is built to handle tasks that used to be impossible.

What Makes Claude Opus 4.6 Different?

Claude Opus 4.6 sits at the top of Anthropic's three-tier system (Opus for complex work, Sonnet for everyday tasks, Haiku for speed). The headline feature? A massive 1 million token context window now available at standard pricing—no more premium charges for long documents . That means you can feed it an entire codebase, thousands of pages of contracts, or the complete transcript of a multi-day meeting, and it will remember everything.

But raw capacity isn't everything. Opus 4.6 also introduces adaptive thinking. Unlike older models where you had to manually turn "extended thinking" on or off, this model decides for itself how much reasoning a task requires . Simple question? You get a fast, cheap answer. Complex coding problem? It silently shifts into high gear, working through edge cases you hadn't even considered. You can also manually set the effort level (low, medium, high, max) to fine-tune the balance between intelligence, speed, and cost.

Key Features That Set It Apart

Agent Teams: Parallel Work, Not Sequential

This is genuinely new architecture. Instead of one AI agent working through tasks one by one, Opus 4.6 can spin up multiple agents that tackle different parts of a problem simultaneously . Imagine one agent reviewing authentication code while another examines database queries and a third checks API endpoints—all at the same time. Early access partner Rakuten reported that "Claude Opus 4.6 autonomously closed 13 issues and assigned 12 issues to the right team members in a single day, managing a ~50-person organization across 6 repositories."

Massive Context That Actually Works

Older models suffered from "context rot"—their performance degraded as conversations grew longer. Opus 4.6 fixes that. On the MRCR v2 benchmark (which tests how well a model retrieves information hidden across massive amounts of text), Opus 4.6 scored 78.3%, the highest among frontier models at that context length . For comparison, its predecessor scored just 18.5% on a similar test . That's the difference between an AI that gets lost on page 200 and one that remembers a tiny detail from page 950.

Context Compaction for Endless Conversations

Ever hit the context limit in the middle of a multi-day project and had to start fresh? Opus 4.6's context compaction automatically summarizes earlier parts of the conversation as you approach the limit, preserving key information while freeing up space . You can work on the same project across multiple sessions without constantly re-explaining context.

Massive Output Length

The model can generate up to 128,000 output tokens in a single response—roughly 96,000 words . That means complete documentation, full codebase migrations, or comprehensive reports delivered in one go.

How People Are Using It

Developers are the obvious winners here. Opus 4.6 achieves state-of-the-art scores on Terminal-Bench 2.0 (65.4%), a real-world coding evaluation that requires planning, error recovery, and context retention . It also leads all other frontier models on Humanity's Last Exam, a complex multidisciplinary reasoning test .

But it's not just for coders. Financial analysts are using it to run complex analyses across hundreds of pages of reports. Legal teams are reviewing contracts with far higher accuracy—Opus 4.6 achieved the highest BigLaw Bench score of any Claude model at 90.2% . Researchers are synthesizing insights from massive document collections. And businesses are automating workflows that previously required teams of humans.

For everyday work, the model can handle spreadsheets, presentations, and research tasks. There's even a research preview of Claude in PowerPoint for automated slide generation and data visualization .

Pricing and Availability

Claude Opus 4.6 is available now on claude.ai, the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry . Pricing remains the same as its predecessor: $5 per million input tokens and $25 per million output tokens . For time-sensitive work, there's a Fast Mode that costs more ($30/$150 per million tokens) but generates outputs up to 2.5x faster .

If you're a Claude Pro, Team, or Enterprise user, you can select Opus 4.6 from the model dropdown on claude.ai. Developers should update their API calls to use model claude-opus-4-6 and migrate from the old thinking parameter to the new adaptive thinking format .

The Bottom Line

Claude Opus 4.6 is the most capable AI model Anthropic has ever released. It's not for simple question-answering—that's what Haiku and Sonnet are for. But if you need an AI that can reason through gnarly problems, navigate massive codebases, and sustain complex tasks over hours or days, this is the model you've been waiting for.