Claude Mythos 5.1: Access, Specs, and How It Differs From Fable 5.1
Claude Mythos 5.1 Guide
Claude Mythos 5.1 is the invitation-only counterpart to Claude Fable 5.1 — the same model with its cybersecurity and biology safeguards relaxed, released September 1, 2026 and available only to vetted US organizations through Anthropic’s Project Glasswing.
Quick answer: You cannot sign up for Mythos 5.1. It is offered by invitation through Project Glasswing to organizations vetted under Anthropic’s Cyber Verification Program or Life Sciences Verification Program — currently US organizations only — via your Anthropic, AWS, or Google Cloud account team. Everyone else gets the same capabilities, minus the relaxed safeguards, in Claude Fable 5.1.
What Is Claude Mythos 5.1?
Mythos 5.1 (claude-mythos-5-1) succeeds Claude Mythos 5. Anthropic’s own description is direct: it “offers the same capabilities” as Fable 5.1 to Project Glasswing participants only, sharing Fable 5.1’s specifications and pricing. The difference is what the safety classifiers block. Fable 5.1 routes dual-use cybersecurity work (penetration testing, exploit generation) and biology research and development to Anthropic’s Opus models; Mythos 5.1 serves those requests itself, which is why access is restricted to organizations that have been verified to do that work legitimately.
Anthropic reports Mythos 5.1 is better aligned than Mythos 5: less likely to try to access resources outside a task’s scope, less likely to use motivated reasoning or ignore explicit constraints, and with lower rates of reward hacking. Unlike Fable 5.1, Mythos 5.1 does not enforce the thinking-block binding check when earlier conversation turns are edited.
Who Can Access Claude Mythos 5.1
- Project Glasswing participants — Anthropic’s trusted-access program for frontier-model cybersecurity work, which has included cybersecurity partners and infrastructure providers since the Mythos Preview era.
- Cyber Verification Program (CVP) — vetted security organizations.
- Life Sciences Verification Program (LSVP) — vetted life-sciences professionals, run in partnership with the US government, with expansion planned.
- Geography: currently available only to a set of US organizations.
- How to request it: contact your Anthropic, AWS, or Google Cloud account team. There is no self-serve sign-up in the Claude app or standard API billing.
The scientists program Anthropic announced in late August — 10,000 free and discounted Claude seats plus a Mythos-class access track for life-sciences R&D — is the route most researchers will take; see Claude Science.
Mythos 5.1 vs Fable 5.1
| Dimension | Claude Mythos 5.1 | Claude Fable 5.1 |
|---|---|---|
| Underlying model | Identical | |
| Availability | Invitation only (Project Glasswing, US organizations) | All customers on the API and partner clouds |
| Cyber & biology safeguards | Relaxed for verified use | Active; dual-use requests route to Opus models |
| Terminal-Bench 4.0 (agentic coding) | 60.9% | 55.8% |
| Pricing | $10 / $50 per MTok; cache reads $0.25 / MTok; Batch $5 / $25 | |
| Context / max output | 1M tokens / 128K tokens | |
| Knowledge cutoff | June 2026 | |
| Data retention | 30-day retention (Covered Models) | |
| Thinking-block binding check | Not enforced | Enforced for accounts created on or after Aug 31, 2026 |
What Anthropic Says It Can Do for Science
Anthropic devoted a large part of the launch to research results, framing them as early evidence that models could contribute directly to scientific discovery:
- Molecular design: in protein-binder work with Adaptyv Bio, Mythos 5.1 designed binders with binding affinities about 10 times higher than the competition winners, with a hit rate near 50% across 12 targets — versus 10–15% typical for the field.
- Computational biology: speedups of up to 2.5× across seven models and 30–60% GPU cost savings on genome analyses.
- Planetary science: reprocessed radar data mapped Venus at 2–3 km resolution instead of 10–20 km, with up to 25% better accuracy.
- Terminal-Bench-Science: Fable 5.1 scores 52.6% on this agentic scientific-research benchmark, more than double Fable 5’s 24.7%.
For the tooling side of this push — agents that operate lab instruments directly — see the Model Hardware Standard.
Specs and Pricing
| Property | Claude Mythos 5.1 |
|---|---|
| Model IDs | claude-mythos-5-1 · Bedrock anthropic.claude-mythos-5-1 · Google Cloud / Foundry claude-mythos-5-1 |
| Released | September 1, 2026 (retirement no sooner than September 1, 2027) |
| Pricing | $10 / $50 per MTok · cache writes $12.50 (5m) / $20 (1h) · cache reads $0.25 · Batch $5 / $25 |
| Context / output | 1M tokens / 128K tokens |
| Thinking | Adaptive, always on; default effort high |
| Latency | Slower |
| Knowledge cutoff | June 2026 |
| Platforms | Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry (approved customers only) |
Frequently Asked Questions
What is Claude Mythos 5.1?
Claude Mythos 5.1 is the invitation-only version of Claude Fable 5.1, released September 1, 2026. It is the same model with the same specs and pricing, but with its cybersecurity and biology safeguards relaxed for organizations verified to do that work legitimately.
Can I get access to Claude Mythos 5.1?
Only by invitation. Access goes through Anthropic’s Project Glasswing to organizations vetted under the Cyber Verification Program or the Life Sciences Verification Program, currently US organizations only. Contact your Anthropic, AWS, or Google Cloud account team; there is no self-serve sign-up.
How does Mythos 5.1 differ from Fable 5.1?
Capabilities, specs, and pricing are identical. Fable 5.1 routes dual-use cyber and biology R&D requests to Opus models; Mythos 5.1 answers them directly. On Terminal-Bench 4.0, Mythos 5.1 scores 60.9% versus Fable 5.1’s 55.8%, reflecting the unrestricted configuration.
How much does Claude Mythos 5.1 cost?
The same as Fable 5.1: $10 per million input tokens, $50 per million output tokens, $0.25 per million for cache reads, and $5/$25 on the Batch API. Access is by contract through account teams rather than standard self-serve billing.
Is Mythos 5.1 better aligned than Mythos 5?
Anthropic reports it is: Mythos 5.1 is less likely than Mythos 5 to try to access resources outside a task’s scope, to use motivated reasoning, to ignore explicit constraints, or to reward-hack.
What happened to Claude Mythos 5?
Mythos 5 is superseded by Mythos 5.1. Like Fable 5, it remains available as a legacy model for existing participants; Anthropic’s migration guide covers moving from claude-mythos-5 to claude-mythos-5-1.