You get a faster Claude that writes better, and there's nothing to install. It's already in your model menu.
Anthropic shipped Claude Opus 5.5 on Tuesday, September 22. It's live in the Claude apps on Pro, Max, Team and Enterprise, and in the API as claude-opus-5-5. Anthropic's own line is that it "performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5."
That's the whole story in one sentence: top-tier work on the model you can afford to use all day. Here's what changed, what to try first, and what to ignore.
What's new, in plain English
Faster. Over 30% faster output than Opus 5. You feel it most on long answers.
More room. Anthropic raised the five-hour usage caps on Pro, Max, Team and Enterprise. Reports put the bump at around 25%. Subscribers also got a one-off limit bonus you can save and reset on a heavy day.
Clearer writing. It puts the key point first, uses less jargon, and sticks closer to the writing rules you give it.
Better at reading charts. Anthropic's docs say it reads charts and diagrams from an image much more accurately than Opus 5, without extra tools.
Cheaper to build on. $4 in and $20 out per million tokens, down from $5 and $25. Cache reads dropped 60%.
Now a quick word from today's sponsor.
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
This guide distills 10 AI strategies from industry leaders that are transforming marketing.
Learn how HubSpot's engineering team achieved 15-20% productivity gains with AI
Learn how AI-driven emails achieved 94% higher conversion rates
Discover 7 ways to enhance your marketing strategy with AI.
Right. Back to it.
Three things to try first
1. Give it the job you'd normally split up
The launch stories are all about size. One team audited and fixed a 200,000-line codebase in under three hours, a job that took more than 20 on Opus 5. Another moved 680,000 lines of code in under a day.
You don't need code to use that stamina. It works the same on a 40-page contract, a year of invoices or a messy spreadsheet export.
Here's [the contract / the spreadsheet / the report]. Read all of it before you answer. Then give me the five things I most need to know, anything that contradicts itself, and anything missing that I'd expect to see. Quote the exact line for every point.The last sentence does the work. With the quote next to each point, you can check it in seconds instead of rereading the whole thing.
2. Paste the chart, not the numbers
Screenshot the dashboard, the sales graph, the slide someone sent you. Older Claude models misread dense charts unless they had tools to zoom in. 5.5 gets most of them right straight from the picture.
This is a screenshot of [what it is]. Tell me what it shows in three sentences, the one number that should worry me, and what I'd check to be sure. If anything is too small to read, say so instead of guessing.3. Hand it your writing rules again
This is where I'd spend the first ten minutes. If you gave up on style rules because Claude drifted back to its own voice after two paragraphs, try again. Following writing guidelines more closely is one of the changes Anthropic named in the launch.
My writing rules: [start with the answer / short sentences / no jargon / words I never use: list them]. Rewrite this in my style and keep every fact: [paste text]What to skip
The benchmark tables. Anthropic itself says benchmark margins "are becoming a less reliable guide". Your own work is the test.
Waiting for Sonnet 5.5 and Haiku 5.5. They're due "in the coming weeks". Opus 5.5 is the one worth using now.
Swapping the model name blind, if you build on the API. Thinking can't be switched off anymore, and forcing a specific tool call now returns an error. Read the migration notes first.
One opinion, and people will argue with me on it. If you've been reaching for Fable for everyday work, stop. Opus 5.5 does that work faster and leaves more room in your limits. Keep Fable for the one job Opus actually gets wrong, and you'll find there are fewer of those than you think.
Make it stick
Put your writing rules where Claude reads them every time. For one area of work, paste them into a Claude Project's instructions. For every chat, put them in Settings, in the personal preferences box. Write them once, and every Opus 5.5 answer starts in your voice.
That's one model upgrade, and you got it for free. A company run by one person gets every upgrade like this across every job at once, as long as the setup's already in place.
The One-Person Company Playbook is the complete build: the folder that holds your business, the brief that runs it, thirty AI roles on staff, six jobs that run on a clock, and the path to your first paying customers. Works on Claude, ChatGPT, Gemini or Grok, whichever you already pay for. About eight hours over a weekend to set up. $97, one-time, updated when the tools change.
Quick one.
What's the first job you gave Opus 5.5? Hit reply, I'm collecting them.
Talk soon,
Zephyr
