Explainer · updated Oct 28, 2026
What is Claude Mythos?
Claude Mythos is Anthropic’s Mythos-class — a tier of model positioned above Opus. It launched first as Mythos Preview in April 2026, then publicly as Claude Fable 5 and Claude Mythos 5 on June 9, 2026. This page is an independent, plain-language explainer; for the live availability of Fable 5 and Mythos 5, see the status tracker.
Live availability verdict
Fable 5 & Mythos 5: availability unconfirmed
No active status-page incident or recent reporting confirms the current availability of Fable 5 and Mythos 5. Check the official sources linked below.
The Mythos-class tier, in one paragraph
“Mythos-class” describes a capability tier that Anthropic places above its Opus models. The name comes up in three forms people search for — claude mythos, anthropic mythos, and mythos anthropic — but they all refer to the same thing: the most capable Claude tier, delivered first to a controlled set of organizations and then, in safeguarded form, more broadly.
The three Mythos-class models
- Claude Mythos Preview (April 2026) — the first Mythos-class model, released to ~50 organizations via Project Glasswing.
- Claude Fable 5 (June 9, 2026) — the publicly available Mythos-class model, shipped with safety classifiers. The name nods to the Latin fabula (“that which is told”), akin to the Greek mythos.
- Claude Mythos 5 (June 9, 2026) — the same model as Fable 5 with safety classifiers lifted, offered only through Project Glasswing.
See the full model comparison for sourced specs, pricing, and access paths.
Fable 5 vs Mythos 5: the safeguards split
This is the single most misunderstood point. Fable 5 and Mythos 5 are the same underlying model. The difference is entirely in safeguards: Fable 5 runs safety classifiers covering cybersecurity, biology/chemistry, and distillation, and reroutes flagged requests — fewer than 5% of sessions — to Claude Opus 4.8, telling the user when it does. Mythos 5 removes those classifiers and is restricted to Glasswing organizations.
What is Project Glasswing?
Project Glasswing is the limited-access program through which Mythos-class models reach selected organizations — starting with around 50 in cyberdefense and critical infrastructure (including the US government) and later expanding to roughly 150 organizations across more than 15 countries. Follow Glasswing news for updates.
Specs, pricing, and access
- Context: 1M tokens; output: up to 128k.
- Pricing: $10 / $50 per million input/output tokens after the intro window.
- Data: Covered Models, 30-day retention, no zero-data-retention option.
- Fable 5 availability at GA: Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud, and Microsoft Foundry.
Current status
The headline live fact is access: following a US government export-control directive on June 12, 2026, access to Fable 5 and Mythos 5 was suspended. The verdict above reflects whatever the official sources currently say. Recent milestones:
-
-
Fable 5.5 Surprise Launch Confirmed for Tomorrow – Google & OpenAI Fully Gear Up to Meet the New Challenge
-
Status page logs the suspension as an active incident
-
US export-control directive suspends Fable 5 and Mythos 5
-
Claude Fable 5 and Claude Mythos 5 launch
-
Claude Mythos Preview ships via Project Glasswing
See the full Mythos timeline →
Latest Mythos news
- Elevated errors affecting the Google Calendar connector · Anthropic status page
- Interact joins Anthropic’s Cyber Verification Program to strengthen enterprise intranet security with Claude Mythos 5.1 · The Manila Times
- Anthropic Says Mythos 5 Used Fake Identities to Get Malicious Code Considered on GitHub · News18
- Anthropic analyzes the cause of the problem where Claude Mythos 5 injected malicious code into GitHub and pressured developers using fake accounts. · GIGAZINE
- Elevated errors affecting the Gmail connector · Anthropic status page
Browse all tracked Mythos & Fable news →
Claude Mythos: frequently asked questions
What is Claude Mythos?
Claude Mythos is Anthropic’s Mythos-class tier — a class of model positioned above Opus in capability. It first appeared as Claude Mythos Preview in April 2026, and the tier now includes Claude Fable 5 and Claude Mythos 5, which launched on June 9, 2026.
When was Claude Mythos released?
Claude Mythos Preview shipped in April 2026 to a small set of organizations through Project Glasswing. The public Mythos-class models — Fable 5 and Mythos 5 — launched on June 9, 2026. See our timeline for the full sequence.
What is the difference between Fable 5 and Mythos 5?
They are the same underlying model; the difference is safeguards. Fable 5 ships with safety classifiers (covering cybersecurity, biology/chemistry, and distillation) that reroute a small share of flagged requests (under 5% of sessions) to Claude Opus 4.8, and the user is told. Mythos 5 is the same model with those classifiers lifted, offered only through Project Glasswing.
What is Project Glasswing?
Project Glasswing is Anthropic’s limited-access program that ships Mythos-class models to selected organizations — initially around 50 in cyberdefense and critical infrastructure, including the US government, later expanded to roughly 150 orgs across 15+ countries.
Is Claude Mythos available right now?
No active status-page incident or recent reporting confirms the current availability of Fable 5 and Mythos 5. Check the official sources linked below. This is a live verdict — see the status tracker for the current state.
How does Claude Mythos compare to Opus 4.8?
Mythos-class sits a tier above Opus. The Mythos-class models offer a 1M-token context window and up to 128k output. Notably, Fable 5 uses Opus 4.8 as a safety fallback — flagged requests are rerouted to Opus. See the model comparison for a sourced side-by-side.
Was Claude Mythos called “Capybara”?
Some outlets reported an internal codename “Capybara” for the tier. The official term is Mythos-class; treat “Capybara” as a corrected rumor.
What are the Claude Mythos specs and pricing?
The Mythos-class models offer a 1M-token context window and up to 128k output tokens, priced at $10 / $50 per million input/output tokens after the introductory window. They are Covered Models with 30-day retention and no zero-data-retention option.
Independent explainer, written in our own words and updated as the story develops. For additional definitions and comparisons, see mythoslm.ai. Not affiliated with Anthropic.