# Mythos-class

> **TL;DR.** Mythos-class is Anthropic's capability tier designation for Claude models that sit above the Opus class — distinguished not just by performance but by a qualitative leap in autonomous cybersecurity, long-horizon reasoning, and scientific hypothesis generation.

- **Category:** AI / Frontier Models / AI Safety
- **Stage:** validating
- **Age:** 66 days
- **Origin date:** 2026-04-07
- **First detected:** 2026-06-12
- **Canonical URL:** https://earlyterms.com/term/mythos-class
- **Sources:** 7 primary URLs

## Definition

Mythos-class is Anthropic's capability tier designation for Claude models that sit above the Opus class — distinguished not just by performance but by a qualitative leap in autonomous cybersecurity, long-horizon reasoning, and scientific hypothesis generation.

The term entered use on [April 7, 2026](https://www.anthropic.com/glasswing) when Anthropic unveiled Claude Mythos Preview alongside Project Glasswing, a consortium of twelve tech giants using the model for defensive vulnerability research. On June 9, 2026, Anthropic released Claude Fable 5 — the first Mythos-class model for general public access — followed by the restricted-access [Claude Mythos 5](https://www.anthropic.com/news/claude-fable-5-mythos-5), available only to vetted cybersecurity and biomedical partners.

## Example

Stripe reported that Claude Fable 5 compressed "months of engineering into days." In parallel, researchers using Claude Mythos 5 generated 14 protein design candidates for drug-target binding — nine proved viable in lab testing — while the same underlying model autonomously discovered 271 buried vulnerabilities in Firefox's codebase.

## Analogy

Think of it as the difference between a skilled locksmith and a model that can redesign the lock industry.

## Why it's emerging now

Anthropic coined 'Mythos-class' to mark a qualitative capability leap — models that autonomously find and exploit zero-days. The June 9, 2026 launch of Fable 5 and Mythos 5 put the term into enterprise discourse, while a mandatory 30-day data-retention policy has made it a live governance battleground.

## Related terms

- *child:* Claude Mythos Preview
- *child:* Claude Fable 5
- *related:* Claude Opus 4.8
- *related:* Project Glasswing
- *parent:* frontier AI
- *related:* ASL (AI Safety Level)
- *related:* agentic AI
- *child:* claude-mythos-preview
- *child:* claude-fable-5
- *related:* claude-opus-4-8
- *related:* managed-agents

## Sources

1. [Anthropic — Project Glasswing announcement (April 7, 2026)](https://www.anthropic.com/glasswing)
2. [Anthropic — Claude Fable 5 and Mythos 5 release (June 9, 2026)](https://www.anthropic.com/news/claude-fable-5-mythos-5)
3. [UK AI Security Institute — evaluation of Mythos Preview cybersecurity capabilities (April 13, 2026)](https://www.aisi.gov.uk/blog/our-evaluation-of-claude-mythos-previews-cyber-capabilities)
4. [Anthropic Help Center — data retention policy for Mythos-class models](https://support.claude.com/en/articles/15425996-data-retention-practices-for-mythos-class-models)
5. [SiliconANGLE — Anthropic sets AI performance records with Mythos 5, Fable 5 (June 9, 2026)](https://siliconangle.com/2026/06/09/anthropic-sets-ai-performance-records-new-mythos-5-fable-5-frontier-models/)
6. [Fortune — Anthropic releases first Mythos-class model to the public (June 9, 2026)](https://fortune.com/2026/06/09/anthropic-releases-its-first-mythos-model-to-the-public/)
7. [Hacker News — 30-day data retention thread (594 points, 298 comments)](https://news.ycombinator.com/item?id=48464258)

---
_Generated by EarlyTerms · https://earlyterms.com/term/mythos-class_