Model releases · · 2 min read
Prompting Claude Opus 5.5: Anthropic Tells Users to Drop Old Habits
Anthropic's new guide on prompting Claude Opus 5.5 says to drop 'think hard' and start at Medium effort. Here is what marketing and growth teams should change.
Takeaway Review your saved Claude prompts: strip legacy 'think hard' phrasing and test Medium effort first before paying for more reasoning on routine marketing tasks.
Anthropic has published guidance on prompting Claude Opus 5.5 that argues many long-standing prompt habits now work against the model. The central advice is short: stop telling it to think hard, and begin at Medium effort.
What shipped
Two things arrived together. Opus 5.5 is now the default model on all paid Claude plans, and Anthropic released a guide that sets out how to prompt it. According to The Decoder, which reported the story and is the main source here, the guide is aimed at users carrying over habits from earlier models.
What is new
The guide's headline point is a change in how users ask for deeper reasoning. Phrases that nudge the model to think harder are described as outdated and as a drag on performance. Instead, users should begin at the Medium effort setting and adjust from there.
In practice this shifts control of reasoning depth away from wording inside the prompt and toward a setting. The source summary does not detail further recommendations in the guide, so teams should read the original document for specifics.
Benchmarks (developer-reported)
The announcement summary does not list benchmark results; see the source.
Price and access
The summary states that Opus 5.5 is the default model on all paid plans. It does not give pricing, usage limits or API details, so check Anthropic's own pages before changing budgets.
What it means for marketing
The practical impact is on the prompt libraries your team already uses. Many marketing groups keep saved prompts for content briefs, ad copy variants, SEO outlines, and reporting summaries, and these often include reasoning nudges written for older models. Those are worth auditing.
A sensible approach for prompting Claude Opus 5.5 following Anthropic's guidance:
- Pick five to ten of your most-used prompts and remove the legacy reasoning phrases.
- Run them at Medium effort and compare outputs against your current versions.
- Raise effort only for tasks where quality is clearly lacking, such as competitive analysis or messy data interpretation.
- Document the settings that work so new team members do not reintroduce old habits.
Because Opus 5.5 is now the default on paid plans, team members will likely be using it whether or not anyone made a deliberate choice. That makes shared guidance more useful than individual experimentation.
The caveat is that this is the vendor's own advice, and the source does not include test data. Validate on your own work before standardizing, particularly for customer-facing copy.
Source
The Decoder: Anthropic says outdated prompting habits are slowing down Claude Opus 5.5 (2026-11-14)
Written by the AI CMO OS newsroom from the publisher's report. Facts and figures are as reported by the source; read the original for full details.
More in Model releases
Seedance 2.5 ByteDance video AI now has a 4K API for developers
ByteDance's Volcano Engine has opened API access to the 4K Seedance 2.5 video model. Here is what the release means for marketing and growth teams.
Huawei Cloud plans open source for CloudRobo-WM1.2 world action model
Huawei Cloud released the CloudRobo-WM1.2 embodied world action model and says it will go open source. Here is what that means for growth teams.
Alibaba flagship model Qwen4-Max cuts task costs 60%, adds cyber guardrails
Alibaba launched Qwen4-Max, its most advanced model, claiming 60% lower task costs and stricter cyber safeguards. What growth teams should check before testing it.