Claude Opus 5.5 arrives
Claude Opus 5.5 begins Anthropic’s new model family, with an emphasis on complex coding and knowledge work. The release pairs capability claims with external evaluation and additional safeguards.
Complex work still needs a review boundary
The Opus 5.5 announcement describes testing by external evaluators, including METR, and stronger results on Anthropic’s automated behavioral audit. It also reports lower running costs than Opus 5 and improvements on demanding software tasks.
Complex work still needs review
- Report boundary
- The announcement’s evaluations do not establish success on an unrelated production task.
View data
| Evidence | Meaning |
|---|---|
| Report boundary | The announcement’s evaluations do not establish success on an unrelated production task. |
Anthropic · published 2026-09-22. Source-bound illustration, not a performance benchmark.
Download imageAnthropic discusses safeguards for biology and cybersecurity, with verified access for particular research uses. The company’s tests and customer examples are useful release evidence, but do not demonstrate that the model will succeed on an unrelated production task or eliminate the need for controls.
Original sources
- Anthropic: original announcementwww.anthropic.com
Checked 9 Oct 2026 · A manually curated edition. Availability may change; company performance claims are not Trion test results. Editorial method.