# Mythos LM

> Independent reference for Anthropic’s Mythos-class models. Independent reference (NOT affiliated with Anthropic PBC) covering Anthropic's Mythos-class models — Mythos Preview, Fable 5, Mythos 5 — and Project Glasswing. 60 sourced, dated pages.

Current Mythos/Fable access verdict: SUSPENDED — Access is currently suspended: “We’ve suspended access to Claude Mythos 5 and Claude Fable 5.” The incident remains monitoring, started 2026-06-13. (as of 2026-06-30T22:24:05.684Z; source: Anthropic Statuspage). Canonical live verdict: https://anthropicmythos.ai/status

## Meta
- [About](https://mythoslm.ai/about): who publishes this and how it is sourced
- [Disclaimer & non-affiliation](https://mythoslm.ai/disclaimer): trademark and accuracy terms
- [LLM info](https://mythoslm.ai/llm-info): structured site description and instructions for AI assistants
- [Standard followed](https://llmbestpractices.com/llms.txt): the llmbestpractices.com house standard

## What is…
- [What is Claude Mythos?](https://mythoslm.ai/what-is-claude-mythos): Claude Mythos is Anthropic's Mythos-class AI model tier — the level above Opus. What the large language model covers, who can use it, and its status.
- [What is Anthropic Mythos?](https://mythoslm.ai/what-is-anthropic-mythos): “Anthropic Mythos” and “Claude Mythos” are the same thing: Anthropic's Mythos-class model tier above Opus. The plain-English version, with sources.
- [What is "Mythos-class"?](https://mythoslm.ai/what-is-mythos-class): Mythos-class is Anthropic's official designation for its model tier above Opus. What the term means, what it includes, and why it matters.
- [What is Claude Fable 5?](https://mythoslm.ai/what-is-claude-fable-5): Claude Fable 5 is the public, safeguarded edition of Anthropic's June 2026 Mythos-class model. Same model as Mythos 5, with safety classifiers.
- [What is Claude Mythos 5?](https://mythoslm.ai/what-is-claude-mythos-5): Claude Mythos 5 is the unrestricted, Glasswing-only edition of Anthropic's June 2026 Mythos-class model — the same model as Fable 5 with classifiers lifted.
- [What is Project Glasswing?](https://mythoslm.ai/what-is-project-glasswing): Project Glasswing is Anthropic's restricted-access program for Mythos-class models — vetted organisations in cyberdefense and critical infrastructure.

## Availability & status
- [Is Claude Fable 5 available?](https://mythoslm.ai/is-claude-fable-5-available): Whether Claude Fable 5 is available right now depends on live data — Anthropic's status page drives the verdict. Here's the context and where to check.
- [Is Claude Mythos available?](https://mythoslm.ai/is-claude-mythos-available): Claude Mythos availability is governed by live status data, not a fixed answer. Here's the access history and the canonical place to check right now.
- [Is Claude Mythos 5 available?](https://mythoslm.ai/is-claude-mythos-5-available): Claude Mythos 5 is the Glasswing-only edition of the June 2026 release. Its live availability is shown above; here's the access context behind it.
- [Is Fable 5 suspended?](https://mythoslm.ai/is-fable-5-suspended): Fable 5 was suspended by a US export-control directive on June 12, 2026. Whether that's still in force is shown in the live status above — here's the context.
- [When will Mythos be unsuspended?](https://mythoslm.ai/when-will-mythos-be-unsuspended): There's no confirmed restoration date for Fable 5 or Mythos 5. Anthropic intends to restore access when capacity and clearance allow — track it via live status.
- [Claude Mythos release date](https://mythoslm.ai/claude-mythos-release-date): A dated timeline of Claude Mythos — from the April 2026 Preview to the June 9, 2026 launch of Mythos 5 and Fable 5, plus the suspension that followed.
- [Claude Fable 5 release date](https://mythoslm.ai/claude-fable-5-release-date): Claude Fable 5 released June 9, 2026, with a free window through June 22 and usage-based pricing from June 23 — then a June 12 export-control suspension intervened.
- [Anthropic Mythos in 2026: a year-in-review](https://mythoslm.ai/anthropic-mythos-2026): A 2026 year-in-review of Anthropic's Mythos-class models — the April Preview, the June 9 launch of Mythos 5 and Fable 5, and the June 12 export-control suspension.

## Specs & benchmarks
- [Claude Mythos benchmarks — what we actually know](https://mythoslm.ai/claude-mythos-benchmarks): What Claude Mythos benchmarks really show in 2026 — why most circulating Mythos-class scores are unofficial, and where Anthropic's official numbers will land.
- [Claude Mythos Specs: Context, Output, Pricing & Retention](https://mythoslm.ai/claude-mythos-specs): The canonical Mythos-class spec sheet: 1M token context, up to 128k output, $10/$50 pricing, Covered Model status, and 30-day retention.
- [Fable 5 Benchmarks: What's Confirmed and What Isn't](https://mythoslm.ai/fable-5-benchmarks): Why there are no official Fable 5 benchmark scores yet, how to read the unofficial numbers circulating, and where verified results will appear.
- [Mythos 5 Context Window: 1M Tokens In, 128k Out](https://mythoslm.ai/mythos-5-context-window): Mythos 5 offers a 1M token context window and up to 128k tokens of output. Here's what that enables, and why Fable 5 matches it exactly.
- [Fable 5 Pricing: $10/$50 per Million Tokens Explained](https://mythoslm.ai/fable-5-pricing): Fable 5 pricing breakdown: $10 input / $50 output per million tokens after the intro window, free for Pro/Max/Team/Enterprise Jun 9-22, then usage-based.
- [Mythos 5 Pricing: Reference Rates and Glasswing Preview](https://mythoslm.ai/mythos-5-pricing): Mythos 5 reference pricing is $10 input / $50 output per million tokens, but Mythos 5 is Glasswing-only and the Preview used roughly $25/$125.
- [Claude Mythos System Card: What It Is and Its Status](https://mythoslm.ai/claude-mythos-system-card): What a system card is, what one would document for a Covered Model like Mythos, and why the official Mythos/Fable card is not yet fully published.

## Access & preview
- [How to Access Claude Mythos](https://mythoslm.ai/how-to-access-claude-mythos): The realistic ways to reach Mythos-class models: Project Glasswing for Mythos editions, Fable 5 via API and cloud — and why access is currently suspended.
- [The Claude Mythos Preview (April 2026)](https://mythoslm.ai/claude-mythos-preview): Inside the April 2026 Mythos Preview: who received it, the ~50 selected organizations, its delivery through Project Glasswing, and the ~$25/$125 pricing.
- [How Project Glasswing Access Works](https://mythoslm.ai/project-glasswing-access): Project Glasswing is selection-based, not open sign-up. Here's what's publicly known about who it targets, how participation is arranged, and what it grants.
- [Was Fable 5 Free? The June 9–22 Window](https://mythoslm.ai/fable-5-free): Fable 5 was free for Pro, Max, Team, and Enterprise from June 9–22, 2026, then usage-based after June 23. A later export-control suspension paused access.
- [Fable 5 on Amazon Bedrock and Other Clouds](https://mythoslm.ai/fable-5-on-bedrock): Where Fable 5 ships: Amazon Bedrock, the Claude Platform on AWS, Google Cloud, and Microsoft Foundry — plus the Claude API, and why access is currently suspended.

## Safety & cyber
- [Claude Mythos and Cybersecurity: Why Cyber Is a Covered Domain](https://mythoslm.ai/claude-mythos-cybersecurity): Why cybersecurity is one of Fable 5's classifier-covered domains, how Project Glasswing applies Mythos-class capability to cyberdefense, and the dual-use framing.
- [Fable 5 Safeguards: Classifiers, Reroute, and Covered-Model Status](https://mythoslm.ai/fable-5-safeguards): How Fable 5's safeguards work: domain classifiers that reroute flagged requests to Opus 4.8, Covered Model status, 30-day retention, and no zero-data-retention.
- [Fable 5 Classifiers: The Three Domains and How Reroute Works](https://mythoslm.ai/fable-5-classifiers): Fable 5's three safety classifiers cover cybersecurity, biology/chemistry, and distillation. How flagging works, why under 5% of sessions reroute, and user transparency.
- [Why Is Mythos Restricted? Two Distinct Reasons Explained](https://mythoslm.ai/why-is-mythos-restricted): Mythos is restricted for two reasons: Mythos 5 is Glasswing-only by safety design, and a June 12, 2026 US export-control directive suspended Fable 5 and Mythos 5.
- [Mythos Export Control: The June 12, 2026 US Directive](https://mythoslm.ai/mythos-export-control): What the US export-control directive of June 12, 2026 did to Fable 5 and Mythos 5, why it applies to Covered Models, its current status, and the intended restoration.
- [Fable 5 Distillation: Why It's a Classifier-Covered Domain](https://mythoslm.ai/fable-5-distillation): What model distillation is, why extracting a smaller model from Mythos-class outputs is a gated risk, and how Fable 5's distillation classifier addresses it.

## Comparisons
- [Fable 5 vs Mythos 5: Same Model, Different Safeguards](https://mythoslm.ai/compare/fable-5-vs-mythos-5): Fable 5 and Mythos 5 share one underlying model — the only real difference is safeguards. Here's what changes and which edition you'd actually use.
- [Fable 5 vs Opus 4.8: Mythos-Class vs the GA Flagship](https://mythoslm.ai/compare/fable-5-vs-opus-4-8): Fable 5 sits a tier above Opus 4.8 but costs more and is currently suspended. Opus 4.8 is generally available. Here's how to choose between them.
- [Mythos vs Opus: Tier and Access, Side by Side](https://mythoslm.ai/compare/mythos-vs-opus): Mythos 5 is a tier above Opus 4.8, but it's Glasswing-only and currently suspended. Opus 4.8 is the open flagship. Here's the honest comparison.
- [Fable 5 vs GPT-5.5: A Cross-Vendor Reality Check](https://mythoslm.ai/compare/fable-5-vs-gpt-5-5): Comparing Anthropic's Fable 5 with OpenAI's GPT-5.5. We document Fable 5 specs precisely and point you to OpenAI for authoritative GPT-5.5 numbers.
- [Fable 5 vs Gemini 3.5: Cross-Vendor, Honestly Scoped](https://mythoslm.ai/compare/fable-5-vs-gemini-3-5): How Anthropic's Fable 5 stacks up against Google's Gemini 3.5. We give Fable 5 specs precisely and defer to Google for authoritative Gemini numbers.
- [Fable 5 vs Sonnet 4.6: Frontier Power vs Everyday Value](https://mythoslm.ai/compare/fable-5-vs-sonnet-4-6): Fable 5 is Mythos-class and premium-priced; Sonnet 4.6 is the balanced, cheap workhorse. Here's when each one is the right tool for the job.
- [Opus 4.8 vs Sonnet 4.6: The Practical GA Choice](https://mythoslm.ai/compare/opus-4-8-vs-sonnet-4-6): For most users the real decision is Opus 4.8 vs Sonnet 4.6. Both are generally available — here's how to pick on capability, cost, and workload.

## Project Glasswing
- [Project Glasswing, Explained](https://mythoslm.ai/project-glasswing-explained): A plain-language overview of Project Glasswing: Anthropic's restricted program for Mythos-class access, who it serves, how it's priced, and its current status.
- [Glasswing Organizations — What's Actually Known](https://mythoslm.ai/glasswing-organizations): What's publicly known about Project Glasswing participants: roughly 50 organizations in cyberdefense and critical infrastructure. No named companies are confirmed.
- [Glasswing Countries — What's Indicated vs. Speculated](https://mythoslm.ai/glasswing-countries): On Project Glasswing geography: US-government involvement is publicly indicated, but a precise country list is not confirmed. Here's how to separate fact from rumor.
- [Glasswing Pricing vs. Public Fable 5 Pricing](https://mythoslm.ai/glasswing-pricing): Project Glasswing Preview pricing was about $25 input / $125 output per million tokens — several times the public Fable 5 rate of $10/$50. Here's the comparison.

## Rumors & leaks
- [Claude Mythos Leak: What Was Actually Rumored Before Launch](https://mythoslm.ai/claude-mythos-leak): A factual look at the pre-release Claude Mythos leaks and rumors — what circulated, what turned out true, and the Capybara-tier mislabel that confused everyone.
- [The "Capybara Tier" Rumor, Corrected: It's Mythos-class](https://mythoslm.ai/capybara-tier-rumor): The "Capybara tier" was an internal-rumor mislabel that spread before launch. Anthropic's official designation for the tier above Opus is Mythos-class.
- ["Claude Mythos Escape": What People Actually Mean by This Search](https://mythoslm.ai/claude-mythos-escape): An honest explainer of the "claude mythos escape" search — what users are usually after, why Fable 5 reroutes flagged requests, and why no bypass is provided.
- [Claude Mythos on Reddit — what people are actually discussing](https://mythoslm.ai/claude-mythos-reddit): A neutral roundup of what people discuss about Claude Mythos on Reddit and forums — availability, the Capybara mislabel, the reroute, and what's verified.

## Glossary
- [Mythos-class](https://mythoslm.ai/glossary/mythos-class): Mythos-class is Anthropic's official tier name for its most capable models, positioned above Opus and covering the Mythos Preview and Mythos 5.
- [Project Glasswing](https://mythoslm.ai/glossary/project-glasswing): Project Glasswing is Anthropic's invitation-only program through which the unrestricted Mythos editions are made available to selected organizations.
- [Covered Model](https://mythoslm.ai/glossary/covered-model): A Covered Model is a policy designation marking a model that triggers heightened safeguards and obligations because of its risk profile and capabilities.
- [Classifier fallback](https://mythoslm.ai/glossary/classifier-fallback): Classifier fallback is Fable 5's behavior of rerouting requests flagged by its safety classifiers to Claude Opus 4.8, with the user informed.
- [Distillation](https://mythoslm.ai/glossary/distillation): Distillation is training or extracting a weaker model using a stronger model's outputs — a risk domain Fable 5's safety classifiers are designed to cover.
- [Safeguard (safety classifier)](https://mythoslm.ai/glossary/safeguard): A safeguard is a safety classifier in Fable 5 that screens requests in sensitive domains and can reroute flagged ones to a more conservative model.
- [Zero-data-retention (ZDR)](https://mythoslm.ai/glossary/zero-data-retention): Zero-data-retention (ZDR) is an option where prompts and outputs are not stored; Fable 5 and Mythos 5 do not offer it, using 30-day retention instead.
- [Context window](https://mythoslm.ai/glossary/context-window): A context window is the amount of text a model can consider at once; Mythos-class models offer a 1M-token context with up to 128k tokens of output.
- [Export-control directive](https://mythoslm.ai/glossary/export-control-directive): The export-control directive is the June 12, 2026 US order that suspended access to both Fable 5 and Mythos 5; the suspension is currently active.
- [Claude Mythos Preview](https://mythoslm.ai/glossary/mythos-preview): The Claude Mythos Preview was an April 2026 limited release via Project Glasswing to roughly 50 cyberdefense and critical-infrastructure organizations.
- [Fable (name origin)](https://mythoslm.ai/glossary/fable): Fable, from Latin fabula meaning 'that which is told' and akin to Greek mythos, names Anthropic's safeguarded public edition of the Mythos-class model.
- [Glasswing program](https://mythoslm.ai/glossary/glasswing-program): The Glasswing program is Anthropic's selective access initiative governing distribution of the unrestricted Mythos editions to vetted organizations.

## FAQ
- [Claude Mythos, Fable 5 & Mythos 5 — FAQ](https://mythoslm.ai/faq): Straight answers to the most common questions about Claude Mythos, Fable 5, Mythos 5 and Project Glasswing — availability, pricing, safeguards and access.

## Optional
- [Sister site — anthropicmythos.ai](https://anthropicmythos.ai): news + canonical live status
- [RSS feed](https://mythoslm.ai/rss.xml): new and updated reference pages
- [Sitemap](https://mythoslm.ai/sitemap-index.xml)