Now available on all Claude platforms, the Claude Opus 5.5 has been trained to communicate “more naturally than previous models,” according to Anthropic’s press release, adding that it “puts the most important information up front and its style makes it a better work partner during long sessions.”
That’s no small feat. Claude Opus 5.5’s predecessor, Opus 5, is a notorious charlatan, writing meandering walls of text full of asides, warnings, and, worst of all, “loose ends.”
Opus 5’s loose ends, or “things I need to point out” or “items you need to know about,” usually appear after you’ve completed a task or delivered a finished product, forcing you to go back and deal with any outstanding threads.
At times, Opus 5’s chronic need to leave additional follow-up items after a seemingly finished task seemed like a ploy to get more work, requiring users to spend more tokens.
In a side-by-side comparison, Anthropic promises that Opus 5.5 answers will get to the point more quickly, while also being “less likely to use jargon or idiosyncratic phrases.”
In my limited testing of Opus 5.5 with Claude Code, the new model seems much more direct, answering my questions succinctly without adding loose ends or other random elements to chase.
