On June 9, 2026, Anthropic launched Claude Fable 5 and Claude Mythos 5. The names look like two separate models, but they are two safety configurations of the same underlying Mythos-class model. Fable 5 is the public version you can call through the Claude API. Mythos 5 is the same base model with some safeguards lifted for a small set of trusted partners.
The practical answer for most developers is simple: use Fable 5. You almost certainly cannot use Mythos 5, and there is no public claude-mythos-5 endpoint to call. The useful part is understanding where the guardrails sit, because that affects how you test prompts, design API integrations, and handle edge-case requests.
TL;DR
Claude Fable 5 and Claude Mythos 5 share the same underlying Mythos-class model.
- Fable 5: public, safeguarded, available through the API and subscription plans.
- Mythos 5: restricted, same base model, some safeguards lifted for vetted users.
-
Pricing: identical at
$10per million input tokens and$50per million output tokens. -
API model id for most developers:
claude-fable-5.
Same model, two safety tiers
Fable 5 and Mythos 5 are not two different models. They are one Mythos-class model exposed under two different safety postures.
Anthropic uses “Mythos-class” to describe the capability tier of the shared model. It is not a separate product name. Both Fable 5 and Mythos 5 inherit that same capability level. For more background, see this breakdown of what Mythos-class means.
The distinction is safeguards:
- Claude Fable 5 is the Mythos-class model configured for public use. It includes built-in protections in three sensitive areas and is available through the Claude API and subscription plans.
- Claude Mythos 5 is the same model with safeguards lifted in some areas. It is not public and is distributed only through controlled trusted-access programs.
For implementation work, this means you should design against Fable 5 unless you are already inside one of Anthropic’s restricted programs.
Claude Fable 5 vs Mythos 5
Use this table as the quick implementation reference.
| Dimension | Claude Fable 5 | Claude Mythos 5 |
|---|---|---|
| Underlying model | Mythos-class, shared | Mythos-class, shared |
| Availability | Public | Not public |
| Safeguards | Built in, active in 3 sensitive areas | Lifted in some areas |
| Who can use it | Anyone with API or subscription access | Small, vetted group |
| Access route |
Claude API using claude-fable-5, plus subscriptions |
Project Glasswing and trusted-access programs |
| Public sign-up | Yes | No |
| Input price | $10 per million tokens | $10 per million tokens |
| Output price | $50 per million tokens | $50 per million tokens |
The model capability is shared. The access model is not. Pricing also stays the same. For more detail on the public version, see this guide to Claude Fable 5 pricing.
How Fable 5 safeguards work
Fable 5’s safeguards are not a blanket downgrade. They are targeted routing rules for specific high-risk categories.
When a request touches one of three sensitive areas, Fable 5 does not answer with the full Mythos-class model. Instead, that request is routed to Claude Opus 4.8, an earlier model with its own safety record. For general background, see this primer on what Claude Opus 4.8 is.
The protected areas are:
Cybersecurity
Requests involving exploit development, offensive cyber operations, or agentic hacking may be routed away from the full model.Biology and chemistry
Requests involving high-risk biological or chemical work, including AAV design and bioweapons-related requests, may be routed.Distillation
Requests that appear designed to extract the model’s behavior to train competing models may be routed.
Anthropic reports that these safeguards trigger in less than 5 percent of sessions on average. For most product workloads — code generation, summarization, extraction, workflow automation, tool use, and general reasoning — you interact with the full Mythos-class model.
The key implementation detail: your endpoint and model id do not change. The routing happens behind the API.
What Mythos 5 unlocks and who gets it
Mythos 5 is the restricted version with some safeguards lifted. It is not available through a public API endpoint or normal sign-up flow.
Known access paths include:
Project Glasswing
Project Glasswing is a collaboration with the US government. Through it, Mythos 5 is provided to cyberdefenders and infrastructure providers with cyber safeguards lifted.
The stated goal is defensive: give vetted teams protecting networks and infrastructure access to the full cyber capability that is not exposed through the public Fable 5 configuration.
Biology researchers
Anthropic has said biology researchers will receive Mythos 5 with biology and chemistry safeguards removed.
That access is scoped. For those users, cyber protections remain active. Mythos 5 is not simply “the model with no rules”; it is the model with specific guardrails lifted for specific vetted use cases.
Broader trusted-access program
Anthropic has also signaled a broader trusted-access program. Public details are limited, but access remains gated rather than self-serve.
Existing Mythos Preview users
Users already inside Claude Mythos Preview can upgrade to Mythos 5. This does not create a public access path for new users.
For most developers, this means Mythos 5 is not part of your architecture decision. Build with Fable 5.
Why Anthropic split the model this way
The split handles a common model-release tradeoff: broad utility versus high-risk capability.
Fable 5 maximizes public availability by routing only the highest-risk categories away from the full model. That keeps most developer use cases on the Mythos-class model while limiting public access to sensitive cyber, bio/chem, and distillation capabilities.
Mythos 5 preserves those capabilities for vetted use cases where the safeguards would block legitimate work. Defensive cyber is the clearest example: if a team is protecting infrastructure, routing cyber queries away from the strongest model may reduce defensive value. So Anthropic lifts that safeguard only for approved users and only through controlled programs.
You can read Anthropic’s own framing in the Claude Fable 5 and Mythos 5 announcement and its broader safety and responsible scaling writing.
What developers should actually do
If you are building with Claude, use:
claude-fable-5
There is no public claude-mythos-5 endpoint.
Your implementation checklist:
- Set the model id to
claude-fable-5. - Budget for
$10per million input tokens. - Budget for
$50per million output tokens. - Test your production prompts against real payloads.
- Check whether any prompts fall near safeguarded domains.
- Document the exact request and response shapes your app depends on.
A typical request body will follow the same Claude API pattern used by other Claude models:
{
"model": "claude-fable-5",
"max_tokens": 1024,
"messages": [
{
"role": "user",
"content": "Summarize this API error log and suggest likely causes."
}
]
}
For request structure and API patterns that carry over to Fable 5, see this Claude Opus 4.8 API guide.
Testing Fable 5 with an API client
When you validate a model integration, do not stop at one successful curl request. Build a repeatable test collection.
A practical workflow:
- Create an environment variable for your Claude API key.
- Add a request that calls the Claude API.
- Set
modeltoclaude-fable-5. - Save representative prompts as separate examples.
- Capture successful and failed responses.
- Share the collection with your team.
- Re-run the same tests before deploying prompt or schema changes.
Apidog can help you send Claude API requests, save them into collections, manage environment variables, and document request/response contracts. That is useful when your team needs reproducible tests instead of one-off commands pasted into chat.
If you are wiring Fable 5 into a larger pipeline, you can also mock and document the surrounding endpoints while the model integration is still changing. Download Apidog to create a Fable 5 request collection, then use Apidog to keep the integration contract visible as your app evolves.
FAQ
Is Claude Mythos 5 the same as Claude Fable 5?
They share the same underlying Mythos-class model. The difference is the safety configuration. Fable 5 has safeguards active in three sensitive areas and is public. Mythos 5 has some safeguards lifted and is restricted to vetted users.
Can I get Claude Mythos 5?
Almost certainly not. There is no public sign-up. Mythos 5 is distributed through controlled programs such as Project Glasswing, an upcoming track for biology researchers, a planned trusted-access program, and upgrades for existing Claude Mythos Preview users.
What does Mythos-class mean?
“Mythos-class” describes the capability tier of the underlying model used by both Fable 5 and Mythos 5. It is not a separate product you can buy. For more detail, see this Mythos-class explainer.
Do Fable 5 and Mythos 5 cost the same?
Yes. Both are priced at $10 per million input tokens and $50 per million output tokens.
What is Project Glasswing?
Project Glasswing is a collaboration between Anthropic and the US government that provides Mythos 5 to cyberdefenders and infrastructure providers with cyber safeguards lifted.
Why does Fable 5 sometimes use Claude Opus 4.8?
Fable 5 routes requests in three sensitive areas to Claude Opus 4.8 instead of answering with the full Mythos-class model. Those areas are cybersecurity, biology/chemistry, and distillation. Anthropic reports that this routing triggers in less than 5 percent of sessions on average.
Where can I see the official model details?
Anthropic’s model overview documentation lists available models and identifiers. The official announcement covers the Fable 5 and Mythos 5 launch. Anthropic also posts policy and safety context on its news and safety pages.
How strong are Fable 5’s safeguards?
An external partner that tested Fable 5 found its cyber safeguards among the most resilient they had evaluated. The safeguards are targeted rather than blanket filters: they route only three sensitive categories away from the full model while leaving other workloads untouched.
Bottom line
For everyday development, build with claude-fable-5.
Fable 5 gives you the public Mythos-class model with targeted safeguards. Mythos 5 is the restricted configuration for vetted programs. Your next step is implementation: add claude-fable-5 to your API request, run real prompts through it, inspect the responses, and document the contract before shipping it to production.
Top comments (0)