Anthropic announced Claude Opus 5.5 on September 22 as the first model in its Claude 5.5 family. Its focus is work that takes several steps: editing code, using a computer, and carrying out professional tasks. For a newcomer, this is a model choice inside Claude or a developer app, not a new kind of AI agent by itself.
What is the improvement?
Anthropic says Opus 5.5 performs near Claude Fable 5.1 on most work while costing 40% less to run than Opus 5. It also reports more than 30% faster output than Opus 5 in its tests. These are the company's own measurements. A benchmark tests a particular task and setup; it cannot tell you how much faster your coding project or document will feel.
Who can use it?
Anthropic says Opus 5.5 is available in Claude for Pro, Max, Team, and Enterprise users, and through its API and major cloud platforms. The API model name is claude-opus-5-5. Anthropic lists standard API prices of $4 per million input tokens and $20 per million output tokens. Those are developer usage prices, not a subscription price. A faster mode costs more.
What should you try?
Give Opus 5.5 one substantial task you can check, such as explaining an unfamiliar codebase and making a small tested change. Compare the result, time, and usage with the Claude model you already use. Anthropic says some sensitive tasks are routed to other models under its safeguards, so the model named in the picker may not handle every request on its own. Its safety report also says pre-release tests cannot catch every failure in real use.