TL;DR
What should I change now that Claude Opus 5.5 is the default?
Anthropic made Claude Opus 5.5 the default model on paid Claude plans and says you can stop writing "think hard" in prompts. Old tricks like that can make replies slower. Delete those lines, leave effort on Medium, and switch to High only when an answer stalls.
What changed in Claude?
Claude Opus 5.5 is now the default model in Claude on the Pro, Max, Ultra and Team plans, Anthropic says. The company recommends it as a "daily driver" for research, analysis, writing, spreadsheets and slide decks. Anthropic hasn't announced any price change alongside it.
Anthropic published a guide on Thursday, November 12, 2026, that explains how to get the most from it. It spread widely over the weekend, mostly because of one tip.
If you're weighing plans, read our Claude Ultra plan explainer.
Why drop "think hard"?
For a while, adding "think hard" or "think carefully" to a prompt was a popular way to get better answers. Anthropic says you can delete those lines now. Opus 5.5 always thinks before it answers and decides on its own how much to think.
Anthropic says that when it removed those phrases, replies started sooner, with no clear drop in quality. That's the company's own finding, so your results may vary, but deleting a few words is an easy thing to try.
How does the effort setting work?
The effort setting controls how much thinking Claude puts into a reply. You'll find it under the model name, next to the send button. Medium is the default on Opus 5.5, and Anthropic suggests starting there and moving to High only when Medium stalls.
Anthropic says Opus 5.5 on Medium "matched or beat" the older Opus 5 on High in its coding and knowledge-work evaluations. Factory's Zimu Li, quoted by Anthropic, said it matched Opus 5 on high effort "while using 20 to 25% fewer output tokens." Output tokens are the chunks of text a model writes back, which is how AI use is measured and billed.
What other prompt habits should change?
The guide has a handful of other tips that work for anyone:
- Say what "done" looks like up front, such as the length, format or the decision you need.
- Attach charts instead of retyping the numbers from them.
- Name the design styles you don't want when asking for slides or layouts.
- Paste in your house style, like a style guide or a sample you like, so Claude matches it.
What developers need to know
Claude Sonnet 4.5 retires on the API on November 30, two weeks from today. The API is the paid connection developers use to build Claude into their own apps and tools. Anthropic lists Sonnet 5.5 as the replacement. Its deprecation page also notes that the temperature, top_p and top_k settings were removed in API version 2026-10-21, so older code may need more than a model name swap.
Our take
Prompt advice ages fast. Plenty of people still paste in tricks from guides written for older models, and "think hard" is the classic example. When the company that makes the model tells you to delete it, that's worth a minute of your time.
The bigger shift is the effort setting. It's a small menu most people will never touch, and that's fine on Medium. Learning to reach for High only on the hard stuff is a simple habit that keeps everyday answers quick.
— Artificially Nerdy
What to do this week
- Clean up your saved prompts. Delete "think hard" and "think carefully" lines.
- Leave effort on Medium and try High only when an answer stalls.
- Add a "done looks like" line to prompts you use often.
- Developers on Sonnet 4.5: move to Sonnet 5.5 before November 30 and check your temperature, top_p and top_k settings.
Still deciding which assistant to pay for? See our ChatGPT vs Claude comparison.
FAQ
Which Claude plans get Opus 5.5 by default?
Pro, Max, Ultra and Team, according to Anthropic.
Does Opus 5.5 cost more than before?
Anthropic hasn't announced a price change with the new default.
What kinds of work is Opus 5.5 meant for?
Anthropic pitches it for research, analysis, writing, spreadsheets and slide decks.
Does the Sonnet 4.5 retirement affect the Claude app?
Anthropic's retirement notice covers the API, the connection developers use. If you only chat in the Claude app on a paid plan, your default is now Opus 5.5.
Sources
- Claude: Using Claude Opus 5.5 as your daily driver for knowledge work — Anthropic, November 12, 2026. Default model, plans, "think hard" advice, effort setting, prompt tips, Factory quote, "matched or beat."
- Claude docs: steering thinking — Medium is the default effort on Opus 5.5.
- Claude docs: model deprecations — Sonnet 4.5 API retirement on November 30, 2026; Sonnet 5.5 replacement; removed temperature, top_p and top_k in API version 2026-10-21.




