Anthropic has announced Claude Fable 5 and Claude Mythos 5, two related models with notably different access and safety arrangements. Fable 5 is described as a Mythos-class model that is safe for general use. Mythos 5 has safeguards lifted for a limited group of cyberdefenders and critical infrastructure providers through Project Glasswing, with broader access planned later.
The distinction matters because the models are not simply two broadly available options at different performance tiers. Anthropic is defining who can use Mythos 5 and for what purpose, while positioning Fable 5 as the model available for general use. The company's official Claude Fable 5 and Mythos 5 announcement also lists pricing and a data-retention policy specific to Mythos-class models.
Importantly, Anthropic's primary materials identify the releases as Claude Fable 5 and Claude Mythos 5. They do not establish a Claude Fable 5.1 or Claude Mythos 5.1 release.
What Anthropic announced
Anthropic describes Fable 5 as a Mythos-class model with safeguards appropriate for general use. That is the key practical point for teams evaluating the announcement: Fable 5 is the generally usable member of this release pair, rather than an access-restricted tool intended only for a narrowly defined security program.
Mythos 5 follows a different path. Anthropic says safeguards have been lifted for a limited set of cyberdefenders and critical infrastructure providers under Project Glasswing. The company has also indicated a pathway to wider availability, but the supplied information does not provide a date or eligibility details for expanded access.
| Area | Claude Fable 5 | Claude Mythos 5 |
|---|---|---|
| General access position | Described as safe for general use | Begins with restricted access |
| Safeguards | Safeguards for general use | Lifted for a limited eligible group |
| Initial intended users | General users | Cyberdefenders and critical infrastructure providers through Project Glasswing |
| Broader availability | General-use positioning | Planned later, with no supplied date |
Pricing and data handling
Anthropic lists pricing of $10 per million input tokens and $50 per million output tokens for the announced models. Token-based pricing means costs depend on both the material sent to a model and the material it generates. For practical planning, a workflow that processes large documents, codebases, logs, or repeated automated tasks can consume substantial input volume even when its final answers are short.
The announcement also specifies a 30-day data-retention policy for Mythos-class models. Organizations considering these models should treat that policy as an implementation factor, especially where prompts may contain operational, technical, or customer information. The supplied research does not provide additional retention terms beyond that 30-day policy, so teams should confirm the current documentation before deploying a sensitive workflow.
What the split means for businesses
For most companies, the immediate relevance is Fable 5's general-use status. A generally available model can be evaluated for everyday knowledge and coding tasks without relying on entry to Project Glasswing. Potential uses depend on a team's own systems and data, but the release creates a clearer separation between routine model adoption and specialized cybersecurity work.
Mythos 5 is more consequential for the specific organizations that qualify for its initial program. Anthropic has tied that access to cyberdefense and critical infrastructure use cases, not to a general-purpose launch. Businesses should therefore avoid treating Mythos 5 as an option they can necessarily procure or deploy today.
A useful evaluation starts with the operational problem, not the model name:
- Identify whether the task is a general knowledge or coding workflow, or a specialized security use case.
- Estimate token usage from the prompts, documents, code, and outputs the workflow will handle.
- Confirm whether the model's access status and 30-day Mythos-class retention policy fit the intended data.
- Keep human review in place for work where incorrect outputs could affect customers, operations, or security decisions.
Advanced models create business value only when they are connected to real work, with clear inputs, review steps, and measurable outcomes. Scalevise helps teams turn model capabilities into dependable processes through practical AI workflow automation, from identifying high-value tasks to connecting tools and reducing manual handoffs. If you want to assess where AI can save time without creating a fragile process, discuss an AI automation project with Scalevise today.
Frequently Asked Questions
What are Claude Fable 5 and Claude Mythos 5?
They are two models announced by Anthropic. Fable 5 is described as safe for general use, while Mythos 5 has a restricted initial access program for certain cyberdefenders and critical infrastructure providers.
Is Claude Mythos 5 generally available?
No. Anthropic says Mythos 5 begins with restricted access through Project Glasswing, with broader access planned later. The supplied information does not provide a date for wider availability.
How much do Claude Fable 5 and Mythos 5 cost?
Anthropic lists pricing of $10 per million input tokens and $50 per million output tokens for the announced models.
What is the data-retention policy for Mythos-class models?
Anthropic states that Mythos-class models have a 30-day data-retention policy.
Did Anthropic announce Claude Fable 5.1 or Claude Mythos 5.1?
No. The official materials supplied for this article refer to Claude Fable 5 and Claude Mythos 5, without a .1 suffix.
Conclusion
Anthropic's announcement is best understood as a split release: Claude Fable 5 is positioned for general use, while Claude Mythos 5 starts with restricted access for defined cyberdefense and critical infrastructure users. Pricing and the Mythos-class retention policy give prospective users concrete factors to assess, but access status remains the central difference between the two models.
Top comments (0)