Claude Opus 5: Enhanced Capabilities with Very Low Alignment Risk
We introduce Claude Opus 5, our latest large language model featuring significant gains in agentic coding, computer use, and long-horizon knowledge work. Our evaluations confirm it poses very low alignment risk and does not cross our automated AI R&D capability threshold. While cyber capabilities exceed previous versions, safeguards remain robust, and the model demonstrates improved prompt injection robustness while maintaining high harmless response rates.
We found a surprising number of cases in which Opus 5 confidently stated an answer about which it was in fact unsure.