Anthropic Releases Claude Opus 5 with New System Card and Benchmarks
Anthropic has officially released Claude Opus 5, its latest flagship large language model, alongside its system card detailing safety and performance metrics. This release provides enterprises with a highly capable model that does not impose strict data retention requirements, addressing critical privacy concerns for corporate adoption. While Opus 5 shows improvements on most benchmarks, discussions have emerged regarding discrepancies in OSWorld 2.0 scoring, where partial credit significantly boosts reported performance compared to full task completion rates.
## BACKGROUND
An AI system card is a document that outlines the design, intended use, limitations, and safety evaluations of an AI model. Claude is a suite of large language models developed by Anthropic, designed to compete with other leading AI systems in reasoning, coding, and general capabilities.