Anthropic says its new Claude Opus 4.6 can nail your work deliverables on the first try

Anthropic / Elyse Betters Picaro / ZDNET

Follow ZDNET: Add us as a preferred source on Google.

ZDNET's key takeaways

• Anthropic debuts Claude Opus 4.6 for enterprise knowledge work.
• It's built for end-to-end autonomy with fewer rewrites.
• Previews include PowerPoint, agent teams, and 1M context.

Anthropic today announced Claude Opus 4.6, which the company says is its most capable model for enterprise and knowledge work. This new large language model is an upgrade to Opus 4.5, with broader autonomy and more accurate first-try results.

Also: Claude Code made an astonishing $1B in 6 months - and my own AI-coded iPhone app shows why

Anthropic describes Opus 4.6 as a "frontier model" designed to handle complex end-to-end enterprise workflows. The term "frontier model" is used by the AI industry to describe AI systems that are at the leading edge of current AI capabilities.

Using Opus 4.6, "Documents, spreadsheets, and presentations will need less back-and-forth on iterations," according to an email ZDNET received from a company representative.

Performance leap for knowledge work

Anthropic says, "For AI to truly tackle enterprise work, it must succeed at three key outcomes: finding information, analyzing it, and producing something from it." According to the company, 4.6 performs well across all three key outcomes.

All this indicates a jump in the AI's agentic capabilities, with an ability to handle complex, long-run tasks in addition to isolated subtasks.

Using travel as an analogy, a simple subtask might be telling a driver to "turn right at the next light," while a more complex task would be to tell that driver located in New York City to drive to Faneuil Hall in Boston. It would be up to the driver to determine the steps and get there. Likewise, the idea with Opus 4.6's broader autonomy is that it can plan and execute the complex series of steps for larger-scale assignments.

Also: How to install and configure Claude Code, step by step

According to the company, Opus 4.6 also reduces the number of corrections and reframes required for "common enterprise deliverables."

According to Yashodha Bhavnani, head of AI at cloud storage vendor Box, "Claude Opus 4.6 excels in high-reasoning tasks, like multi-source analysis, across legal, financial, and technical content. Box's eval showed a 10% lift in performance, reaching 68% vs. a 58% baseline, and near-perfect scores in technical domains."

Anthropic is also positioning Claude Opus 4.6 as a valuable resource for financial modeling. The AI can help with regulatory filings, market reports, and internal data, producing rapid results for projects that would previously take analysts days to complete. Anthropic says Opus 4.6 "handles the nuance required for compliance-sensitive output."

Opus 4.6 is proving to be powerful for legal reasoning as well. According to Niko Grupen, head of AI research at legal AI company Harvey, "Claude Opus 4.6 achieved the highest BigLaw Bench score of any Claude model at 90.2%. With 40% perfect scores and 84% above 0.8, it's remarkably capable for legal reasoning."

Another intriguing new capability is Claude's integration with PowerPoint. Once released, Claude will be able to work directly inside PowerPoint (presumably as a plugin) and be able to read layouts, fonts, and slide masters. This way, edits by the AI can stay "on-brand and on-template."

Also: I tried a Claude Code alternative that's local, open source, and completely free - how it works

According to the company, Claude Opus 4.6 can "build slides from a corporate template, restructure a storyline, convert bullets into diagrams, or generate a full deck from a description -- all without leaving the app."

The PowerPoint capability is in research preview, available via a waitlist. ZDNET has requested access. As soon as we get it, we'll create some spiffy slides and report back to you.

Developer and agent advances

Claude is particularly well known for its agentic coding capabilities. Claude Opus 4.6 builds on the strengths of Opus 4.5 with more agentic behavior. The company says that autonomous coding improvements will particularly benefit developers with large code bases, long-horizon tasks, and complex implementations.

Also: Stop using ChatGPT for everything: My go-to AI models for research, coding, and more (and which I avoid)

As a user of Claude Code, this brings to mind a key question. Claude Code using Opus 4.5 often needs to run compaction sequences that free up available resources. This process not only takes a long time, but it often interrupts project flow.

If 4.6 is supposed to be able to address even larger code bases, then the context window needs to grow. Anthropic says that "Claude Opus 4.6 will support 1M context (in beta) at launch. This is the first Opus model with long context." It'll be very interesting to see that in action.

Agent teams

The company is offering a research preview of agent teams in Claude Opus 4.6 to API and subscription Claude users. The company says teams "let Claude Code work the way a real engineering team does. Instead of one agent working through tasks sequentially, you can split the work across multiple agents -- each owning its piece and coordinating directly with the others."

Also: I let Anthropic's Claude Cowork loose on my files, and it was both brilliant and scary

I've been struggling with Claude running multiple parallel agents in Claude Code using Opus 4.5, particularly in the Xcode 26.3 preview. I've found that once the primary agent kicks off a series of subagents, they're not visible for my hands-on management. When one or more of them gets stuck (as they seem to do with disturbing regularity), the whole agentic coding process just hangs.

I'm hoping that agent teams in Claude Opus 4.6 provide better transparency, better overall management, and better damage control, so if they get stuck, they report back and ask for help. Stay tuned. I'll do some testing and report back on overall performance.

That said, Michele Catasta, president of AI no-code company Replit says, "Claude Opus 4.6 is a huge leap for agentic planning. It breaks complex tasks into independent subtasks, runs tools and subagents in parallel, and identifies blockers with real precision."

Availability

Anthropic says, "Claude Opus 4.6 is available today on claude.ai, our API, and all major cloud platforms." Token pricing hasn't changed from the previous release for API users. 

Some features like PowerPoint, the 1M context, and agent teams are described as research previews or beta, and are not available for wide release at launch. But Anthropic is working on AI time. So items in research preview and beta are more likely to be weeks away than months away. After all, it does have an AI to help it code its products.

Also: Want local vibe coding? This AI stack replaces Claude Code and Codex - and it's free

What do you think about Claude Opus 4.6 and Anthropic's push toward more autonomous, enterprise-focused AI? Do you see real value in features like agent teams, 1M context, or deep integrations like PowerPoint? Would you trust an AI to handle complex work end-to-end with less human oversight, or do you still prefer tighter control? How do you think this compares to other frontier models you've used? What questions do you still have about availability or real-world performance? Let us know in the comments below.

You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.