Claude Fable 5 vs Claude Mythos 5: What Is the Difference?
Fable 5 and Mythos 5 share the same underlying model but serve very different audiences. Here is how they differ and who gets access to each.
TL;DR
Claude Fable 5 and Claude Mythos 5, both announced by Anthropic on June 9, 2026, are the same underlying model. Fable 5 is the publicly available version with full safeguards - the first Mythos-class model made safe for general use. Mythos 5 lifts safeguards in some areas and is restricted to cyberdefenders and infrastructure providers through Project Glasswing, plus select biomedical researchers later via a trusted access program.
One Model, Two Releases
Anthropic's June 9 announcement introduced a deliberate split. Rather than choosing between releasing a maximally capable model and a maximally safe one, the company shipped both - with access controls doing the work that capability limits used to do.
- Claude Fable 5: state-of-the-art on nearly all tested capability benchmarks, especially long and complex tasks, available to everyone on the Claude API, Amazon Bedrock, and GitHub Copilot
- Claude Mythos 5: the same model with safeguards lifted in some areas, carrying the strongest cybersecurity capabilities of any model in the world
How Fable 5 Stays Safe
Fable 5's safeguards are layered. Cybersecurity classifiers monitor queries, and when one triggers, the query is answered by Claude Opus 4.8 instead. Biology and chemistry classifiers cover other high-risk domains, and distillation prevention protects the model itself. The friction is modest: safeguards trigger in under 5 percent of sessions on average, and over 1,000 hours of external red-teaming found no universal jailbreaks. Anthropic's alignment assessment found misaligned behavior at levels similar to Opus 4.8.
One operational note for businesses: Mythos-class models require 30-day data retention for business customer traffic, used for safety monitoring only, not training.
Who Gets Claude Mythos 5
Mythos 5 access runs through Project Glasswing, a collaboration with the US government that upgrades the earlier Claude Mythos Preview deployment and expands it to roughly 150 new organizations across more than 15 countries. Participants are cyberdefenders and infrastructure providers. Later, select biomedical researchers will gain access through a trusted access program.
The early science results explain why researchers want it. Internal protein-design experts accelerated drug design roughly 10x, with 9 of 14 protein targets yielding strong drug-design candidates. In blind comparisons, molecular biologists preferred its hypotheses about 80 percent of the time over Opus-class models, and one E. coli protein hypothesis was independently corroborated.
Which One Matters for You?
| Question | Fable 5 | Mythos 5 |
| Who can use it | Anyone on Claude plans or the API | Vetted organizations only |
| Safeguards | Full classifiers and fallbacks | Lifted in some areas |
| Pricing | 10 and 50 dollars per million tokens | Restricted program access |
| Best for | Coding, agents, general work | Cyberdefense, biomedical research |
For nearly all developers and businesses, Fable 5 is the answer - it is the same intelligence, packaged for general use. Mythos 5 exists for the narrow set of defenders and researchers whose work justifies lifted safeguards.