tl;dr: Anthropic released Claude Opus 5, claiming state-of-the-art performance on coding and knowledge-work benchmarks (Frontier-Bench, ARC-AGI 3, OSWorld 2.0) at half the cost of their top "Fable 5" model, though it trails "Mythos 5" on cybersecurity exploitation tasks. Priced at $5/$25 per million input/output tokens (same as Opus 4.8), it emphasizes improved judgment, self-verification, and agentic reliability, and is now the default on Claude Max. Safety testing rates it their most aligned model yet, with relaxed cyber classifiers and optional automatic fallbacks for flagged requests.