Anthropic launches Opus 5, a cheaper, less restricted AI model
Anthropic has released Opus 5, a cheaper and less restricted AI model that outperforms its flagship Fable 5 on key benchmarks, offering European businesses a more flexible enterprise tool without strict data retention policies.
Anthropic released Opus 5 on Friday, adding a new option to a rapidly expanding AI lineup that has seen Mythos 5, Fable 5, and Sonnet 5 all launch in June. The lightweight Haiku model remains the only holdout waiting for an upgrade to the 5 series.
Despite being smaller and cheaper than the flagship Fable 5, Opus 5 outperforms it on several benchmarks included in the announcement. Anthropic highlighted that the new model is “much stronger at verifying its work and iterating carefully until it succeeds,” noting it successfully built a computer vision pipeline from an incomplete prompt during testing.
For European companies balancing cost and capability, this combination of a lower price point and high performance makes Opus 5 a likely preferred choice over Fable in most commercial use cases. By releasing a model that undercuts its own premium product on price while matching or exceeding its output, Anthropic is positioning itself to capture a broader share of the enterprise market.
Privacy-conscious firms will likely welcome Opus 5’s data handling. Unlike Fable and Mythos, the new model is not subject to a 30-day data retention policy. This distinction removes a significant point of friction for businesses operating under strict regional data protection rules, giving Anthropic a clearer pitch to heavily regulated industries.
Anthropic has also applied a lighter touch to Opus 5’s safety classifiers, expecting them to trigger 85% less often than those on Fable 5. The model still retains specific guardrails against offensive cybersecurity tasks. It blocks vulnerability scans on compiled software binaries but permits them in source code, a restriction Anthropic bases on the defensive nature of the latter task.
To prevent workflow disruptions for developers, Anthropic is introducing an Automatic Fallbacks beta feature for API users. When a prompt triggers a safety block, the system will automatically route the request to a less powerful model. This ensures companies receive a functional response rather than a hard error message, reducing the operational friction that often accompanies strict AI moderation tools.
