Anthropic makes Claude Fable 5 public as a safer Mythos-class model

Anthropic has opened public access to Claude Fable 5, presenting the model as a safer version of its higher-end Mythos-class technology. The company said the release is designed to bring stronger reasoning and long-horizon task performance to a broader audience while keeping tighter controls around the areas where frontier models can create risk.

The launch is notable because Anthropic has previously treated its most capable systems as sensitive technology, especially when they can plan across many steps, write code, interpret technical documents, or assist with research workflows. By branding Fable 5 as Mythos-class, the company is signaling that it sits closer to the top of the Claude family than a routine incremental update.

Anthropic makes Claude Fable 5 public as a safer Mythos-class model

A public model with stronger guardrails

The key distinction in the announcement is not simply benchmark performance. Anthropic is positioning Claude Fable 5 as a model that can be used widely without exposing every capability of the underlying Mythos line in every context. For everyday users, that means the model should handle complex writing, analysis, coding, and planning tasks. For more sensitive prompts, the system is expected to apply stricter safety behavior.

That design reflects a broader shift in the AI market. Labs are no longer competing only on raw speed or context length; they are also trying to prove that powerful models can be released in ways that customers, regulators, and enterprise buyers will trust. A public Mythos-class model gives Anthropic a way to compete at the high end while still emphasizing its safety-first brand.

Why the release matters

For developers and companies already building on Claude, Fable 5 could become a new default for demanding tasks that require persistence across documents, codebases, and decisions. It may be especially relevant for agentic workflows where a model is asked to decompose a goal, call tools, revise outputs, and keep track of constraints over time.

The release also raises the stakes for rivals. If users can access a safer public version of technology previously associated with private frontier deployments, pressure will grow on other AI providers to explain how they balance capability, availability, and misuse prevention. Claude Fable 5 therefore arrives not just as a product update, but as a test of whether top-tier AI systems can move into mainstream use without abandoning safety boundaries.