Anthropic makes Claude Opus 5.5 the default and says to drop "think hard"

Anthropic has made Claude Opus 5.5 the default model on its Pro, Max, Ultra and Team plans, according to The Decoder. The company has also shared prompting advice for the model. Its main message: many users hold the new model back with instructions carried over from older generations.

What changes in your prompts

The first habit Anthropic wants gone is adding phrases such as "think hard" or "think carefully." The company says Opus 5.5 always reasons by itself and adjusts how deeply it reasons to the task and to the effort level you choose. In Anthropic's own tests, answers written without those phrases began sooner, with no measurable loss in quality. The Decoder notes these phrases come from the early days of reasoning models, roughly a year and a half ago, before adaptive reasoning became standard.

Anthropic also suggests a different way of asking. Instead of feeding instructions one at a time, put the whole request in one message: the goal, the format you want back, and when the model should stop to ask you questions. The company says Opus 5.5 is better than its predecessor at carrying multi-part tasks to the end, and that it can handle much of the back-and-forth users used to do themselves.

Other tips from Anthropic:

• Attach diagrams and tables as images or files rather than typing them out. The company says workarounds like copying chart values by hand are no longer needed.
• Ask for the finished document or spreadsheet directly, without requesting a draft first.
• For design work, name the specific styles you do not want, such as a cream background or pill-shaped buttons. Anthropic says vague bans on "generic design" just trade one default look for another.
• Share your own house style, and connect Claude to your files and systems through connectors. Anthropic says connected setups produce much more tailored results because the model can fetch what it needs.

Anthropic also says image reading improved a lot. By its account, Opus 5.5 at its lowest effort level read values from dense charts more accurately than Opus 5 at its highest.

What effort level to use

Medium is the default effort setting for Opus 5.5, and Anthropic recommends starting there. The company says that in internal tests, Opus 5.5 on Medium performed as well as Opus 5 on High. Higher settings cost more and do not always give better answers, Anthropic says, so it advises moving to High only when Medium falls short. You can change the setting in the model menu beside the send button.

What it costs

Anthropic says teams that used the cheaper Sonnet for most work and saved Opus for hard projects should reconsider. Its argument is that a more capable model can finish a job in fewer steps, which partly offsets the higher price per token. The company adds that people who often push Sonnet to "Extra High" or "Max" effort end up paying about as much as they would with Opus 5.5. These cost and performance comparisons are Anthropic's own.

What to watch

All of the speed, quality and cost claims here come from Anthropic's internal testing. The Decoder's report does not include independent measurements. Compare your own results on Medium and High before changing how your team works.