Anthropic ships Claude Fable 5 to the public, keeps Mythos 5 locked to cyber defenders
The Mythos-class model goes general availability with classifier guardrails that punt cyber, bio, chem, and distillation prompts to Opus 4.8 — while a sibling build with the cyber safeguards stripped stays inside Project Glasswing.
Anthropic released Claude Fable 5 into general availability on Tuesday alongside Claude Mythos 5, a sibling build available only to vetted users. Both run on the same underlying model. The difference is a layer of safety classifiers, which the company says is “the point: one model, two products, distinguished by a layer of safety classifiers.”
Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, less than half of April’s Mythos Preview and roughly double Claude Opus 4.8. Pro, Max, Team, and seat-based Enterprise plans get it bundled until June 22, after which TechCrunch reports usage credits will be required until capacity lets Anthropic restore it as a standard feature.
The classifiers intercept prompts touching cybersecurity, biology, chemistry, and model distillation, then hand the session off to Opus 4.8. Anthropic says they fire on under 5% of sessions and that it tuned them conservatively to ship quickly. The Hacker News notes the distillation block is also aimed at preventing rivals from extracting Mythos-class capabilities to train competing models.
The reason for all of this becomes legible once you read the Project Glasswing results. During April’s Mythos Preview, the model identified and exploited zero-day vulnerabilities in every major operating system and browser when directed, including a 27-year-old OpenBSD flaw and CVE-2026-4747, a remote-code-execution exploit against FreeBSD’s NFS server. Mythos 5 stays inside Glasswing, accessible to the US government and other defenders. The unsafetied weights don’t leave the building.
Anthropic ran more than 1,000 hours of external bug bounty testing and reports no universal jailbreak on long-form agentic tasks, though The Hacker News flagged that the UK’s AI Security Institute made partial progress inside an early window.
Two other details matter. Fable 5 and Mythos 5 traffic carries a mandatory 30-day retention period, overriding prior zero-retention agreements; TechCrunch suggests this could set an industry precedent. And per CNBC, the launch lands days after Anthropic confidentially filed its IPO prospectus with the SEC against a $47 billion revenue run rate and a $965 billion valuation.
The choreography is the story. The cyber-capable model goes to the government. The retention policy hardens. The S-1 goes to the SEC. CISA, NSA, and Five Eyes guidance on agent deployment is sitting on the same desks.
Sources
- https://www.anthropic.com/news/claude-fable-5-mythos-5
- https://techcrunch.com/2026/06/09/anthropic-released-claude-fable-5-its-most-powerful-model-publicly-days-after-warning-ai-is-getting-too-dangerous/
- https://www.cnbc.com/2026/06/09/anthropic-mythos-claude-fable-5.html
- https://thehackernews.com/2026/06/anthropic-releases-claude-fable-5-its.html
- https://cyberscoop.com/cisa-nsa-five-eyes-guidance-secure-deployment-ai-agents/
— END —