The company Anthropic advises on how to adjust prompts for the Claude Opus 5.5 model
The company Anthropic recommends that users of the Claude Opus 5.5 model abandon old prompts like "think hard", start at the Medium effort level, and specify the goal in a single message — according to the company, this speeds up responses without loss of quality and can also reduce costs.
The company Anthropic has published recommendations on how to adjust the way prompts are written for the Claude Opus 5.5 model, which is now the default on the Pro, Max, Ultra and Team plans. According to the company, the model works differently from its predecessors, and many users are unknowingly holding it back with outdated instructions inherited from older generations of models.
The main recommendation is to omit phrases like "think hard" or "think carefully" from prompts — according to the company, the model reasons automatically and adapts the depth of its reasoning to the task and the chosen effort level. In the company's tests, responses without these phrases started faster with no measurable drop in quality. The company further recommends starting at the default Medium effort level, which in its internal tests achieved performance comparable to the High level of the previous Opus 5 model; switching to High only when Medium is not sufficient. Teams that, for cost reasons, used Sonnet and reserved Opus for complex tasks should, according to the company, reconsider this approach, because a more capable model completes the task in fewer steps, and running Sonnet at Extra High or Max effort ends up costing similarly to Opus 5.5.
Another recommendation is to describe the goal of the task in a single message, including the required output format, instead of giving instructions step by step — according to the company, Opus 5.5 is better at completing multi-task assignments and can effectively break the task down on its own. It is better to attach diagrams and tables as images or files rather than retype their values into text; according to the company, the model reads even densely packed charts more accurately at the lowest effort level than Opus 5 does at the highest level. For final documents or tables, the company recommends asking directly for the finished result, because the model often manages it on the first attempt.
For design tasks, the company recommends specifically naming unwanted styles (for example, a cream-colored background or buttons with rounded corners) instead of a vague ban on "generic design", because according to the company, generic instructions just swap one default style for another. Users who connect the Claude model to their own files and systems via connectors, according to the company, get significantly more accurate results.
Why it matters
The recommendations change how individuals and teams formulate their requests, and can directly affect both the speed of responses and the cost of using the model. Companies that until now saved money by deploying the cheaper Sonnet model at high effort may, according to the company Anthropic, find that Opus 5.5 at a lower effort level ends up costing similarly but completes the task faster.
Two audiences, two different impacts
What this means
For individuals
Users of the Claude Opus 5.5 model should abandon habits from earlier models: not writing phrases like "think hard" (according to the company Anthropic, the model reasons automatically), starting at the Medium effort level instead of High, specifying the goal in a single message, and attaching diagrams or tables as images instead of retyping the values.
For a business
According to the company Anthropic, companies that use Sonnet for cost reasons and reserve Opus only for complex tasks should reconsider this practice: a more capable model can often complete the task in fewer steps, which can offset the higher price per token.
ProductivityCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.