Anthropic tells Claude Opus 5.5 users to drop "think hard" prompts

Anthropic has published prompting advice for Claude Opus 5.5, according to The Decoder. The core message: anyone using the model should first drop several habits carried over from older model generations. Anthropic says Opus 5.5, now the default on Pro, Max, Ultra and Team plans, works differently from its predecessors, and that many users hold it back with outdated instructions.
First on the list is cutting phrases like "think hard" or "think carefully". According to Anthropic, the model always reasons on its own and scales its reasoning depth to the task and the selected effort level. In Anthropic's tests, responses without these phrases started faster with no measurable drop in quality. The Decoder explains that such phrases date back to the early days of reasoning models, before today's standard adaptive reasoning existed. "Back then," about a year and a half ago, users could sometimes coax more performance out of a model or steer the router this way.
Anthropic also recommends starting at the Medium effort level, which is the default for Opus 5.5. In internal tests, Opus 5.5 on Medium matched the performance of its predecessor Opus 5 on High, the company says. Higher settings cost more and don't always produce better results, so Anthropic advises switching to High only when Medium isn't enough. Users can change the setting in the model menu next to the send button.
The company also asks teams that used the cheaper Sonnet and kept Opus for complex projects to rethink that split. Its argument is that a more capable model can finish a task in fewer steps, which can partly offset the higher per-token price. Anyone who regularly runs Sonnet at Extra High or Max effort ends up paying about as much as with Opus 5.5, Anthropic says.
The guidance then moves from chatting to setting goals. Anthropic advises describing the desired result in a single message rather than giving instructions one at a time, covering the goal, the output format, and when the model should check back with questions. Opus 5.5 is better than its predecessor at seeing multi-part tasks through, according to the company. The model can increasingly take over the back-and-forth that users used to handle in the chat, sometimes doing it better, and effectively prompt itself from a single initial input.
On inputs, Anthropic recommends attaching diagrams and tables as images or files instead of typing them out. Workarounds older models required, such as manually transcribing chart values, are no longer needed, the company says. Image reading improved significantly: even at the lowest effort level, Opus 5.5 read values from dense charts more accurately than Opus 5 at the highest, the firm says. Users who need a finished document or spreadsheet should ask for it directly instead of requesting a draft first, since Opus 5.5 often goes from request to finished product in a single pass.
For design tasks, Anthropic recommends naming unwanted styles specifically, such as a cream-colored background or pill-shaped buttons, rather than vaguely banning "generic design". According to the company, general instructions only swap one default style for another. Finally, users who connect Claude to their own files and systems through connectors get significantly more tailored results, Anthropic says, because the model can pull in the information it needs on its own.
Key facts
- Anthropic advises cutting phrases like "think hard" or "think carefully" from Opus 5.5 prompts; in its tests, responses without them started faster with no measurable drop in quality.
- Start at the Medium effort level, the Opus 5.5 default; Anthropic says Opus 5.5 on Medium matched Opus 5 on High in internal tests, and High is for when Medium isn't enough.
- Teams that kept Opus for complex work and used Sonnet otherwise should rethink that: Anthropic says regular Sonnet use at Extra High or Max effort costs about as much as Opus 5.5.
- Describe the goal, output format and when to ask questions in a single message, ask directly for finished documents, and attach charts and tables as images or files.
- For design, name specific unwanted styles (a cream-colored background, pill-shaped buttons) instead of banning "generic design".
Why it matters
Many people write prompts the way they did for earlier models, and Anthropic says that now works against them. The "think hard" trick dates from the early days of reasoning models, about a year and a half ago, when it could sometimes coax more performance out of a model or steer the router. With adaptive reasoning now standard, Anthropic says the model decides how deeply to reason on its own. The practical shift is from coaxing the model with phrases to stating the goal clearly and choosing an effort level.
Who it affects
Anyone using Claude Opus 5.5, which is now the default on Pro, Max, Ultra and Team plans. Teams that split work between Sonnet for routine tasks and Opus for complex projects are addressed directly. So are people who type out chart values or ask for drafts before a final document, and anyone prompting for design work.
How to use it
Remove "think hard" and "think carefully" from prompts. Leave the effort setting at Medium, the default, and switch to High only when Medium falls short; the setting sits in the model menu next to the send button. Describe the whole task in one message: the goal, the output format, and when the model should check back with questions. Attach diagrams and tables as images or files. Ask for the finished document or spreadsheet directly. In design prompts, name the specific styles you don't want. Connect Claude to your own files and systems through connectors so it can pull in what it needs.
How solid is it
This is Anthropic's own guidance as relayed by The Decoder, and nearly every claim is attributed to the company: the faster responses, the Medium-versus-High match, the chart-reading comparison and the cost comparison with Sonnet. No benchmark names, sample sizes or numeric results are given for the internal tests, and the comparisons are qualitative. The article does not say how much faster responses started without "think hard" phrases. No independent verification of Anthropic's claims is reported.
Risks and caveats
Higher effort settings cost more and don't always produce better results, per Anthropic, so a wrong default can waste money. The Sonnet cost argument rests on the company's say-so; no per-token prices for Opus 5.5 or Sonnet are given, and the Sonnet version is not named. The claim that Medium on Opus 5.5 matches High on Opus 5 comes from internal tests only. The article also does not give a publication date for the guidance.
“Higher settings cost more and don't always produce better results.”
— The Decoder, reporting Anthropic's guidance