Anthropic launches Opus 5, a cheaper, less-restricted flagship model

Artificial intelligence company Anthropic launched its Opus 5 model on Friday, the newest version of its long-standing heavyweight system. Though smaller than the company’s Fable 5 model, Opus 5 will be both cheaper and less restrictive than Fable — which Anthropic says likely makes it preferable in most use cases. The company also says Opus 5 actually outperforms Fable 5 on a number of benchmarks included in its announcement.

The release lands fast on the heels of its predecessors: Opus 5 arrives only two months after Opus 4.8, which became available on May 28, and after Mythos 5, Fable 5, and Sonnet 5 all launched in June — leaving only the lightweight Haiku model still waiting for an upgrade to the 5 series. In the post announcing the new model, Anthropic said Opus 5 is “much stronger at verifying its work and iterating carefully until it succeeds,” citing benchmark testing in which the model wrote its own computer-vision pipeline in response to an incomplete prompt, among other examples.

Opus 5 also sheds many of the restrictions that have applied to Fable since its release. Like its predecessor, Opus 5 is not subject to the 30-day data-retention policy that covers Fable and Mythos — a policy that had raised concerns among privacy-conscious users. Meaningful safeguards remain in place, particularly around cybersecurity tasks such as exploit generation and penetration testing: Opus 5’s safeguards prevent it from being used to scan a software binary for vulnerabilities, though it is still permitted to search for vulnerabilities in source code, since that task is more likely to be used for defensive purposes. Broadly, Anthropic expects these safety classifiers to engage 85% less often for Opus 5 than they will for Fable 5, a reflection of the lighter touch given to the less capable model.

Anthropic is also rolling out a new tool meant to make the safeguards less disruptive when they do engage. Users can now opt in to a beta feature called Automatic Fallbacks, which automatically routes a request to a less powerful model whenever a prompt triggers the safety classifier. The result: API users with the setting enabled get a functional response instead of an error message.

Why it matters

By pairing a cheaper, less-restricted flagship model with a fallback system that swaps in a smaller model rather than returning an error, Anthropic is betting that fewer refusals and interruptions matter more to developers building production tools than raw model size.