Claude Fable 5 launch brings Mythos-class intelligence to the public as Anthropic adds safety guardrails and restricts the full Mythos 5 to approved partners

Anthropic made a major announcement. The company launched Claude Fable 5, a Mythos-class AI model now available to the general public. The Claude Fable 5 launch marks the first time Mythos-level capability has been accessible outside Project Glasswing, which previously limited such models to selected cyberdefenders and infrastructure providers. Moreover, Anthropic simultaneously launched Claude Mythos 5 for a smaller group of approved users. Therefore, both models are now live — but with very different levels of access.

Safety Guardrails

Fable 5 uses the same underlying model as Mythos 5 but includes specific safeguards for public use. If the system detects requests involving cybersecurity, biology, chemistry, or model distillation, it hands the response to Claude Opus 4.8 instead. Furthermore, users receive a notification when this fallback occurs. Anthropic acknowledges the guardrails are intentionally cautious and may occasionally block harmless requests. However, fallback happens in fewer than 5 percent of sessions on average.

Claude Mythos 5

Mythos 5 uses the same base model as Fable 5 but removes certain safeguards in specific areas. It is currently available only to Project Glasswing partners in collaboration with the US Government. Furthermore, Anthropic plans to expand access through a trusted-access programme for cybersecurity organisations and a separate biology access programme for selected researchers. Therefore, full Mythos 5 access remains tightly controlled for now.

Performance Claims

Anthropic says Fable 5 outperforms earlier public models across most AI benchmarks. In one example, Stripe used Fable 5 to complete a codebase-wide migration across a 50-million-line Ruby codebase in a single day. Furthermore, Anthropic said the same task would have taken a human team more than two months. The model also performs strongly in coding, finance reasoning, chart interpretation, scientific vision, and long-context work.

Research Uses

Mythos 5 has shown strong results in life sciences. Anthropic’s internal protein design experts used it to speed up parts of the drug design process by around ten times. Furthermore, Mythos 5 produced molecular biology hypotheses that Anthropic scientists preferred over Opus-class outputs approximately 80 percent of the time in blinded comparisons.

Data Retention

Anthropic introduced a 30-day data retention policy for both models. Furthermore, the company confirmed this data will not train new Claude models or serve non-safety purposes. The policy aims to help detect jailbreaks, complex misuse attempts, and false positives.

Pricing and Availability

Claude Fable 5 is available now through the Claude API as claude-fable-5. Both models cost $10 per million input tokens and $50 per million output tokens. Furthermore, Fable 5 is included at no extra cost for Pro, Max, Team, and seat-based Enterprise users through June 22. On June 23, those users will need usage credits to continue accessing it. Finally, Claude Mythos 5 remains restricted to Glasswing partners until the trusted-access programme opens to a wider audience.