Introducing Claude Opus 5

Anthropic has released Claude Opus 5, a new model positioned between Opus 4.8 and the top-tier Fable 5, at the same price as its predecessor ($5/M input, $25/M output tokens). It sets new state-of-the-art results on Frontier-Bench and GDPval-AA coding/knowledge work benchmarks, triples the next-best model on ARC-AGI 3, and outperforms all competitors on OSWorld 2.0 computer use at roughly one-third the cost of Fable 5. Key improvements include stronger agentic behavior (self-verification, iterative debugging, building its own test harnesses), better scientific reasoning across life sciences, and improved visual output quality. It becomes the default model on Claude Max and the strongest on Claude Pro. Safety evaluations show it is Anthropic's most aligned model to date, with the lowest deceptive behavior rates. Two beta features ship alongside it: mid-conversation tool changes without cache invalidation, and automatic model fallbacks when safety classifiers trigger. Fast mode is available at 2.5× default speed for 2× base price.