Introducing Claude Opus 5

Anthropic announced the release of Claude Opus 5, positioning it as a step‑change improvement over its predecessor Opus 4.8. The new model is described as thoughtful and proactive, offering intelligence close to Claude Fable 5 while costing half as much. On coding and knowledge‑work benchmarks such as Frontier‑Bench and GDPval‑AA, Opus 5 is the new state‑of‑the‑art, though it remains behind Mythos 5 on cybersecurity tasks. The model is set as the default on Claude Max and the strongest on Claude Pro, and it delivers improved performance at the same cost as Opus 4.8. Evaluations show that Opus 5 surpasses all other models on Frontier‑Bench v0.1, doubling Opus 4.8’s performance at lower cost per task. On CursorBench 3.2, at max effort, it performs within 0.5% of Fable 5’s peak score while costing half as much. Opus 5 also shows notable gains in scientific research, outperforming Opus 4.8 on life‑sciences evaluations, especially in organic chemistry and protein‑related tasks. Early‑access reports highlight its strong visual outputs, agentic coding, debugging, and thoroughness. The model topped Zapier’s AutomationBench leaderboard, achieving 100% on a churn‑prevention sequence without exceeding token usage of prior Claude models. In genomics analysis, Opus 5 behaves like a careful scientist, selecting appropriate statistical tests and cross‑checking results. Across internal evaluations, Opus 5 is steadier with less variance, and it excels in full‑stack app builds, financial modeling, legal agent work, and visual formatting. Anthropic notes that Opus 5’s judgment and verification capabilities enable it to handle complex tasks with fewer oversight requirements. The company plans to migrate use cases in its Cosmos platform and adopt Opus 5 for production workloads, expressing confidence that users will prefer it over Opus 4.8.