Claude Opus 5.5 prompting: Anthropic says delete "think hard"

Claude Opus 5.5 prompting works differently from older models, and Anthropic now says so directly. In a guide published on Nov 12, 2026, it writes that "each new model is a reason to retest what you hand to Claude and to drop prompting habits built for older models." The first habit on its list is one many people type out of routine: "think hard."
Headlines summed it up as "old habits slow Claude down." Anthropic's own wording is more precise. Without those lines, replies start sooner, with no clear drop in quality. The guide applies to Pro, Max, Ultra and Team plans, where Opus 5.5 is now the default model. The Free plan doesn't have Opus 5.5.
Delete "think hard" and "think carefully"
Anthropic's guide says: "Thinking is always on for Opus 5.5, and it decides how much to think based on your request and the effort level. In our testing, removing those lines made replies start sooner, with no clear drop in quality."
So a prompt that starts with "Think hard about this" spends time on an instruction the model already follows. Take the line out and write the task itself.
Start at Medium effort
Effort is the setting that controls how long Claude works on a reply. Anthropic says Medium is the default for Opus 5.5, and that "on our coding and knowledge-work evaluations, Opus 5.5 at Medium matched or beat Opus 5 at High." Higher settings "think longer and cost more without always producing better work."
The advice is to stay on Medium and move to High only when Medium stalls. To change it in the Claude app, click the model name next to the send button and choose Effort.
Ask for the finished file and say what done looks like
Two more habits on the list come from working with older models step by step:
- Ask for the spreadsheet or document itself, not an outline. Anthropic says that "more often than not, Opus 5.5 can get from your request to the finished file in one pass."
- In your first message, describe the goal, the finished output, and when Claude should stop and ask. The guide says Opus 5.5 "carries long, multipart work through to the end better than Opus 5 did."
Attach the chart instead of retyping it
If you used to copy a chart's numbers into the prompt so the model could read them, share the chart, screenshot or diagram instead. Anthropic calls retyping values one of the "workarounds built for earlier models" that "may no longer be needed."
The guide also suggests naming the exact styles you don't want in design work (it gives "a cream background or pill-shaped buttons" as examples) and pasting your brand or writing guidelines into the request.
FAQ
Should I still write "think step by step" for Claude?
For Opus 5.5, Anthropic says to delete "think hard" and "think carefully" because thinking is always on. In its testing, replies started sooner without them and quality didn't clearly drop.
What effort level should I use for Claude Opus 5.5?
Start at Medium, the default. Anthropic found Opus 5.5 on Medium matched or beat Opus 5 on High. Switch to High only when Medium gets stuck.
Can I use Claude Opus 5.5 on the Free plan?
No. Anthropic lists Opus 5.5 as the default model on Pro, Max, Ultra and Team plans.
Sources
Tools in this article
Anthropic's AI assistant: chat, files, artifacts and the Claude Code app. Free plan; paid plans start with Pro.
Open ClaudeWe earn nothing from these links. All tools
Making something to sell?
Our guide shows how to make and sell one small PDF with AI, with the real numbers. Chapter 1 is free.
Read next
- ChatGPT can now pay for you. Not in regular chatsNov 14, 2026 · AI news
- Gemini 4 Hydrogen is free. The app picks whenNov 13, 2026 · AI news
- Gemini 4 Carbon is out. One catch on Nov 30Nov 12, 2026 · AI news
Checked against official sources on Nov 15, 2026. The character in our posts is made with AI; the text is written with AI help and checked by hand rules.