Anthropic released Claude Opus 5 on Friday, priced at $5 per million input tokens and $25 per million output tokens, roughly half the cost of the company’s own flagship Fable 5. Opus 5 becomes the default model on Claude Max and the strongest option available on Claude Pro.
Anthropic says Opus 5 sets new state-of-the-art marks on its Frontier-Bench and GDPval-AA evaluations while still trailing Mythos 5 on cybersecurity tasks. In other words: near-frontier on knowledge work, one tier down on the specialized capability the company is most conspicuously careful about.
“Enterprises, in our feedback and with our customer base, are looking for value,” Dianne Penn, Anthropic’s head of product management for research, told CNBC. The framing tracks. Opus 5 ships with a user-facing effort dial (low, medium, high) that lets customers trade intelligence against token spend, and an Automatic Fallbacks beta that reroutes requests tripping safety classifiers to a less-restricted model instead of returning an error. Anthropic expects classifiers to engage about 85 percent less often on Opus 5 than on Fable 5.
This is the fourth Claude 5-series model in under two months, following Mythos 5, Fable 5 and Sonnet 5 in June and Opus 4.8 on May 28. Only the lightweight Haiku tier remains un-upgraded.
The cadence has a subtext. In June, Anthropic disabled access to Mythos 5 and Fable 5 for about two weeks under a U.S. export-control directive before restoring both on June 30. Opus 5, the company notes, has been submitted to government partners for independent testing. The state-capacity layer is now part of the product roadmap, not adjacent to it.