Claude Mythos 5 Explained

Restricted as of July 1, 2026: Claude Mythos 5 is not generally available. Access is limited to vetted partners in Anthropic's Project Glasswing program, with a planned expansion to biomedical researchers (Anthropic). This page explains Claude Mythos 5 in plain terms.

Claude Mythos 5 uses the same base model as Claude Fable 5, but with safeguards lifted for authorized users. It is Anthropic's most capable model for high-stakes technical work. Most businesses cannot buy it directly.

This guide covers what Mythos 5 is, why it is restricted, who can access it, and its capabilities and limits. It also explains the compliance terms and why Claude Fable 5 is the practical choice for nearly every company.

Reviewed by Jonathan West, Founder of Layer3 Labs, on July 1, 2026. We research using primary vendor and regulator sources.

What Is Claude Mythos 5?

Claude Mythos 5 is a restricted Anthropic model that uses the same underlying model as Claude Fable 5 but with safeguards lifted for authorized users. The difference is governance, not raw intelligence. Anthropic released it on June 9, 2026.

The lifted safeguards let approved users do high-stakes technical work that the public Fable 5 model blocks. That includes offensive-security research, complex analysis, drug design, and novel hypothesis generation (Anthropic).

Mythos 5 is aimed at a narrow audience. Anthropic names cyberdefenders, infrastructure providers, and biomedical researchers as the intended users. It is not a general office productivity tool.

• Same base model as Claude Fable 5, with safeguards lifted for authorized users.
• Released June 9, 2026 by Anthropic.
• Built for cyberdefenders, infrastructure providers, and biomedical researchers.
• Anthropic published a red-team assessment of its cybersecurity capabilities.

Key Features and Capabilities

Claude Mythos 5 is Anthropic's most capable model for high-stakes technical work. It handles tasks the public Fable 5 model refuses, such as offensive-security research. It shares the deep reasoning strength of the shared base model.

The model is strong on complex analysis, drug design, and new hypothesis generation. These fit its target users in security and biomedical research. Anthropic tested the risks before release.

Anthropic published a red-team assessment of the model's cybersecurity capabilities. That review documents what the model can do and supports the case for tight access. This page keeps to capabilities and governance only.

• Offensive-security research that the public Fable 5 model blocks.
• Complex technical analysis for security and infrastructure work.
• Drug design and novel hypothesis generation for biomedical research.
• A published red-team assessment of its cybersecurity capabilities.

Pricing and Access

Claude Mythos 5 has no public price or self-serve signup because it ships only through a restricted program. Access is limited to vetted partners in Anthropic's Project Glasswing program. A general business account cannot enroll on its own.

Project Glasswing is a restricted program for cyber-defense and infrastructure partners. Anthropic plans a separate trusted-access program for biomedical researchers. Both paths require vetting.

Access for US organizations was restored on June 26, 2026. This followed US government approval after a brief June export-control review. Availability can shift with policy changes.

• No public price list and no self-serve signup.
• Access is gated through Anthropic's Project Glasswing program.
• A planned trusted-access program is set for biomedical researchers.
• US access was restored June 26, 2026 after government approval.

Best Uses for Business

Claude Mythos 5 fits a narrow set of vetted organizations, not general business teams. Its lifted safeguards only help users with a real need for high-stakes technical work. For everyone else, Claude Fable 5 is the right tool.

Cyber-defense teams and infrastructure providers use it for offensive-security research under Project Glasswing. Biomedical researchers are the planned next audience via a trusted-access program. These are specialized roles, not everyday office work.

A common mistake is assuming a technical business qualifies by default. Access depends on vetting and current export-control policy, not on company size or budget. Most companies should plan around Fable 5 instead.

• Cyber-defense teams: offensive-security research through Project Glasswing.
• Infrastructure providers: high-stakes technical analysis under vetting.
• Biomedical researchers: drug design via a planned trusted-access program.
• General business teams: use Claude Fable 5, which is available now.

Limitations and Risk Factors

The biggest limitation of Claude Mythos 5 is that nearly no business can access it. It ships only through a restricted program, so it is out of reach for general teams. This alone rules it out for most use cases.

The lifted safeguards raise real misuse risk. Removing guardrails that block high-risk cyber and bio requests is why access is so tightly gated. Anthropic keeps the user base narrow to manage that risk.

Availability is also tied to policy. US access was restored on June 26, 2026 after an export-control review, but rules can change again. Treat access as conditional, not guaranteed.

• Not generally available; no self-serve path for most businesses.
• Lifted safeguards raise misuse risk, so access stays narrow.
• Availability depends on current export-control policy.
• Compliance terms are not the standard Enterprise defaults.

Compliance and Data Handling

Claude Mythos 5 ships through a restricted program, so its compliance terms are program-specific. The usual Enterprise BAA and zero-retention terms do NOT automatically apply. You must confirm terms inside your Project Glasswing agreement.

Business accounts carry a mandatory 30-day safety retention. That means data is held for safety review, then deleted. Full zero-data-retention is not available on this model.

A common failure mode is assuming standard Enterprise terms carry over. They do not, because Mythos 5 does not use the standard Enterprise surface. Confirm HIPAA and retention terms in writing before sending any regulated data.

• Program-specific compliance terms, not standard Enterprise defaults.
• Mandatory 30-day safety retention on business accounts.
• No zero-data-retention option on this model.
• Confirm HIPAA and retention terms in your program agreement.

How Claude Mythos 5 Compares

For nearly every business, Claude Fable 5 is the right choice and Mythos 5 is out of reach. The two share a base model, so raw capability on normal tasks is close. The real split is access and safety guardrails.

Mythos 5 lifts high-risk cyber and bio guardrails for vetted users and ships through Project Glasswing. Fable 5 keeps those guardrails on and is generally available worldwide as of July 1, 2026. Anthropic reports more than 95% of Fable 5 sessions never trigger a safety fallback.

Choose Mythos 5 only if you are a vetted cyber-defense, infrastructure, or biomedical partner with a real need for lifted safeguards. Choose Fable 5 for standard business work, since it is available now and easier to govern.

• Availability: Mythos 5 is restricted (Project Glasswing); Fable 5 is generally available worldwide.
• Safeguards: Mythos 5 lifts high-risk guardrails for authorized users; Fable 5 keeps them on.
• Compliance: Mythos 5 uses program-specific terms; Fable 5 offers standard Enterprise and API terms plus a HIPAA BAA.
• Verdict: pick Fable 5 unless you qualify for and truly need Mythos 5.

How to Access Claude Mythos 5

You access Claude Mythos 5 only by being vetted into Anthropic's Project Glasswing program. There is no direct purchase or self-serve signup. General businesses have no path to it today.

Cyber-defense and infrastructure partners apply through Project Glasswing and pass Anthropic's vetting. Biomedical researchers will access it through a planned trusted-access program that keeps cyber safeguards on while removing biology safeguards. Both routes stay gated.

If you cannot get Mythos 5, deploy Claude Fable 5 instead. It uses the same base model with safeguards on and is generally available worldwide as of July 1, 2026. It runs on the Claude Platform, Claude.ai, Claude Code, and Cowork.

• Cyber-defense and infrastructure partners: apply through Project Glasswing.
• Biomedical researchers: a planned trusted-access program (cyber safeguards kept).
• General businesses: no direct self-serve path today.
• Deployable alternative: Claude Fable 5, available worldwide now.