ℹ️ Quick Answer: Claude Opus 4.6 launched today with three standout features. “Agent teams” let multiple AI agents split tasks and work in parallel. A 1 million token context window (in beta) can process about 1,500 pages in a single prompt. And Claude now works inside PowerPoint. Pricing stays at $5/$25 per million tokens on the API.
📋 WHAT’S INSIDE
- Agent Teams: Claude Opus 4.6 Splits the Work
- 1 Million Token Context Window
- Claude in PowerPoint
- How Claude Opus 4.6 Stacks Up
- What This Means for You
Anthropic and OpenAI both dropped major announcements today. Claude Opus 4.6 is Anthropic’s new flagship model, and the upgrades go beyond the usual “it’s smarter” claims.
I’ve been using Claude daily for close to a year. So I want to walk through what actually changed and why it matters.
Agent Teams: Claude Opus 4.6 Splits the Work
Agent teams let you spin up multiple Claude instances that tackle different parts of a project at the same time. One reviews code while another writes tests and a third researches documentation.
This is the feature that caught my attention. Instead of one AI agent working through a task step by step, you can now assign a group project. The agents share results and stay in sync, so you’re not manually stitching pieces together afterward.
Right now this is available in Claude Code as a research preview, designed for tasks like codebase reviews that split naturally into independent subtasks. For developers, this is a big deal. For everyone else, it signals where AI assistants are heading. Your future AI won’t just be one helper. It’ll be a whole team.
1 Million Token Context Window
The new 1 million token context window (in beta) processes about 1,500 pages in a single prompt with 76% retrieval accuracy, up from 18.5% in the previous model.
Claude Opus 4.6 can now process about 1,500 pages of text, 30,000 lines of code, or over an hour of video in a single prompt. That’s an entire book worth of information at once.
The context window is in beta, and Anthropic says the model scores 76% accuracy on retrieving specific details from within that massive context. The previous model scored 18.5% on the same test. Not a typo. It also supports up to 128,000 output tokens, so it can generate much longer responses than before.
Claude in PowerPoint
Claude now sits inside PowerPoint as a side panel, reading your existing layouts, fonts, and brand templates to generate slides that actually match your style instead of producing generic decks.
If you build presentations regularly, pay attention to this one. Ask Claude to build a deck and it pulls from your existing layouts, fonts, and templates. The result actually looks like your brand instead of a default AI slide dump.
Available as a research preview for Max, Team, and Enterprise plan users. Anthropic also improved Claude in Excel for longer tasks and unstructured data.
How Claude Opus 4.6 Stacks Up
Opus 4.6 beats GPT-5.2 by 144 Elo points on general reasoning, scores 90.2% on BigLaw Bench for legal tasks, and found over 500 previously unknown vulnerabilities in open source code.
Anthropic says Opus 4.6 beats both GPT-5.2 and Gemini 3 Pro on most benchmarks. The model outperforms GPT-5.2 by about 144 Elo points on general reasoning. On legal tasks (BigLaw Bench), it scored 90.2%. On life sciences, it performed nearly twice as well as the previous Opus.
The security angle is interesting too. Anthropic says Claude Opus 4.6 found over 500 previously unknown vulnerabilities in open source code using just its default capabilities. That’s both impressive and a little unsettling.
What This Means for You
The Opus 4.6 upgrade is automatic and live now on claude.ai for all users, with noticeably better answers on complex questions and the ability to process much longer documents.
If you already use Claude, the upgrade is automatic on claude.ai. It’s live now. You’ll notice better answers on complex questions and the ability to feed it much longer documents.
If you’ve never tried Claude, this is a solid time to start. Our Claude Opus 4.5 review covers how the previous version performed, or visit our Start Here page if you’re new to AI tools.
Leave a Reply