TL;DRAnthropic launched Claude Fable 5, its first publicly available Mythos-class model, with safeguards that block high-risk queries and fall back to Opus 4.8.
Anthropic on Tuesday released Claude Fable 5, a model built on the same architecture as its restricted Mythos system, making Mythos-class intelligence publicly available for the first time. Fable 5 is available to enterprise customers and paid subscribers, but it comes with new safeguards that block responses in cybersecurity, biology, chemistry, and what Anthropic calls “distillation” scenarios, falling back to Claude Opus 4.8 to deliver a safe answer instead. Early data from Anthropic indicates the fallback triggers in fewer than 5% of sessions, meaning most users will interact with the full Fable 5 model for the vast majority of their queries.
“For us, it’s really around what we call ‘race to the top,’ being able to provide this technology in a valuable fashion, and at the same time providing the right safety guardrails so that it can do asymmetrically more benefits than harm,” Dianne Penn, Anthropic’s head of product management for research, told CNBC. The model represents what Penn described as a “significant jump” in capability over Opus 4.8, which is why the additional guardrails were necessary. On some benchmarks, Fable 5 scored more than 10% higher than Opus 4.8, according to Anthropic’s own blog post, though independent evaluations have not yet been published.
