Anthropic's Opus 5 beats Fable 5 on benchmarks, fewer restrictions
Outperforms its bigger sibling while being cheaper and less restrictive.
Anthropic released Opus 5 on Friday, the latest version of its long-standing heavyweight model. Despite being smaller than Fable 5, Opus 5 outperforms Fable 5 on several benchmarks cited in the announcement. It is also cheaper and has fewer restrictions, making it preferable for most use cases. The launch comes just two months after Opus 4.8 (May 28), while Mythos 5, Fable 5, and Sonnet 5 debuted in June — leaving only the lightweight Haiku model awaiting a 5-series upgrade. Anthropic emphasized that Opus 5 is "much stronger at verifying its work and iterating carefully until it succeeds," pointing to benchmark tests where it wrote its own computer vision pipeline from an incomplete prompt.
Crucially, Opus 5 is free from the 30-day data retention policy that applies to Fable and Mythos, addressing privacy concerns. It still has safeguards around cybersecurity tasks (e.g., exploit generation) but triggers them 85% less often than Fable 5. Anthropic also introduced a beta feature called Automatic Fallbacks, which routes requests to a less powerful model when safety classifiers fire, returning a functional response instead of an error. This lighter touch makes Opus 5 more accessible for developers and businesses seeking high performance with fewer friction points.
- Opus 5 outperforms the larger Fable 5 on multiple benchmarks while being cheaper and less restrictive.
- No 30-day data retention policy, unlike Fable and Mythos, addressing privacy concerns for API users.
- New Automatic Fallbacks feature prevents error messages by routing requests to weaker models when safety classifiers activate.
Why It Matters
Developers get better performance and lower costs with fewer safety restrictions, streamlining AI integration into workflows.