Anthropic's most powerful AI ever built — the one they were once too afraid to release. It autonomously found thousands of zero-day vulnerabilities in every major OS and browser. As of June 9, 2026, it's finally public.
On June 9, 2026 Anthropic launched two versions: Claude Mythos 5 (safeguards lifted, restricted to vetted partners) and Claude Fable 5 — the same model with standard safeguards, available to everyone. Fable 5 scores 80.3% on SWE-Bench Pro, the new #1, and is free until June 22.
Claude Mythos is the most capable model Anthropic has ever built. It was accidentally exposed through a data leak in late March 2026 before Anthropic officially announced it on April 7 as a private preview.
In internal testing, Anthropic used Mythos to scan major software infrastructure — and the results alarmed them. Mythos autonomously discovered thousands of zero-day vulnerabilities across every major operating system and every major web browser. Not with human guidance. Autonomously, after a single request to find bugs.
One specific example stands out: Mythos fully autonomously identified and then exploited a 17-year-old remote code execution vulnerability in FreeBSD — a flaw that allowed root access on any machine running NFS. No human was involved after the initial request.
Anthropic initially restricted access to launch partners through Project Glasswing, letting the world's largest software companies find and fix their vulnerabilities before attackers could. Two months later, on June 9, 2026, the company released Mythos publicly — saying broad availability became possible because of new safeguards that block responses in specific high-risk areas. The general-access version ships as Claude Fable 5; the unrestricted version remains Claude Mythos 5, for vetted partners.
Fable 5 and Mythos 5 are the same underlying model. Fable 5 brings the capability to everyone with safety controls that block high-risk cyber responses; Mythos 5 lifts those safeguards for vetted partners doing advanced cybersecurity research. Identical benchmarks, identical price — the only difference is what each is allowed to do.
Anthropic's controlled initiative that used Mythos defensively during its private preview — finding vulnerabilities before attackers do.
These organizations used Claude Mythos to scan their own systems — operating systems, browsers, cloud platforms, financial infrastructure — and patch vulnerabilities before malicious actors could find them. That defensive track record is part of why Anthropic was confident enough to release a safeguarded version, Fable 5, to the public in June.
What you can access today — and how the Mythos-class models compare to the previous flagship.
| Feature | Mythos 5 | Fable 5 | Opus 4.8 |
|---|---|---|---|
| Availability | Vetted partners | Everyone | Everyone |
| Safeguards | Lifted in some areas | Standard (cyber blocked) | Standard |
| SWE-Bench Pro | 80.3% | 80.3% | 69.2% |
| GDPval-AA | 1,932 | 1,932 | 1,890 |
| Input pricing | $10 / 1M | $10 / 1M | $5 / 1M |
| Output pricing | $50 / 1M | $50 / 1M | $25 / 1M |
| Context window | 1M tokens | 1M tokens | 1M tokens |
| Best for | Advanced security research | Hardest agentic + coding work | Best-value premium default |
Bottom line: If you want the most capable model you can use, that's Claude Fable 5. If you want frontier-class coding at half the price, Claude Opus 4.8 still beats every non-Anthropic model. Only vetted partners can access the unrestricted Mythos 5.
Mythos accidentally leaked via a data exposure. Anthropic confirms its existence.
Anthropic announces Claude Mythos Preview and Project Glasswing. 12 partner organizations receive access.
Anthropic announces EU access to its advanced Mythos model, widening the controlled rollout.
Public release: Claude Fable 5 (general availability, Mythos-class) and Claude Mythos 5 (safeguards lifted, partners only). 80.3% SWE-Bench Pro.
End of the free-access window for Fable 5 — API usage billed at $10/$50 per 1M tokens after this date.
Claude Mythos is Anthropic's most capable model line, first announced April 7, 2026 as a private preview (Project Glasswing) because of its extreme cybersecurity capabilities — it can autonomously discover and exploit zero-day vulnerabilities. On June 9, 2026 Anthropic released it publicly in two forms: Claude Mythos 5 (safeguards lifted, restricted to vetted partners) and Claude Fable 5 (the same model with standard safeguards, generally available to everyone).
Yes — through Claude Fable 5. Fable 5 is the same underlying model as Mythos 5 with safeguards that block specific high-risk cyber responses, and it's available to all users and API customers. The unrestricted Mythos 5 (with safeguards lifted) remains gated to vetted enterprise and research partners. For 99% of people, Fable 5 is Mythos — just the version you're allowed to use.
They are the same underlying model with identical benchmarks — 80.3% SWE-Bench Pro and 1932 GDPval-AA. Fable 5 is generally available with safety controls that block high-risk cyber responses. Mythos 5 lifts those safeguards for advanced security research and is restricted to vetted partners. Both are priced at $10/$50 per 1M tokens.
In internal testing, Mythos autonomously discovered thousands of zero-day vulnerabilities across every major operating system and browser — and exploited a 17-year-old RCE in FreeBSD entirely without human involvement. Anthropic concluded a broad release was irresponsible until defensive infrastructure caught up. The June 9 public release became possible because of new safeguards that block responses in specific high-risk areas.
Project Glasswing was Anthropic's initiative to use Claude Mythos defensively during its private preview. Anthropic committed $100M in model usage credits to launch partners — including AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks — to find and fix vulnerabilities in critical shared infrastructure before attackers could exploit them.
Fable 5 (the Mythos-class public model) scores 80.3% on SWE-Bench Pro vs Claude Opus 4.8's 69.2% — an 11-point jump, the biggest single-release gain of 2026. It costs double, though: $10/$50 per 1M tokens vs Opus 4.8's $5/$25. For the hardest agentic work, Fable 5 wins; for everyday volume, Opus 4.8 is the better value.
$10 per million input tokens and $50 per million output tokens — double Claude Opus 4.8. Anthropic made Fable 5 free for all users until June 22, 2026 to drive adoption. It has a 1M-token context window and 128K max output per request.
One notable example: Mythos fully autonomously identified and then exploited a 17-year-old remote code execution vulnerability in FreeBSD — a flaw that allowed root access on any machine running NFS. No human was involved in either the discovery or exploitation after the initial request was made.