---
title: 'Claude Fable'
description: "Anthropic's generally available Claude model line for demanding reasoning and long-horizon work, using safeguards that distinguish its deployment from Mythos."
canonical_url: 'https://darkfactory.dev/glossary/fable'
markdown_url: 'https://darkfactory.dev/glossary/fable.md'
collection: glossary
date_published: '2026-10-07T00:00:00-04:00'
date_modified: '2026-10-07T00:00:00-04:00'
---

# Claude Fable


## Definition

Claude Fable is Anthropic's generally available Claude model line for demanding reasoning and long-horizon work. Its deployment includes safeguards that distinguish it from Mythos. For the 5 and 5.1 releases, Anthropic states that Fable and Mythos share the same underlying model, with different safeguards and access rules.

Fable names a deployed product with an operating policy as well as a model release. The full version and serving environment describe the system a user encounters.

## Origin and attribution

Anthropic launched Claude Fable 5 and Claude Mythos 5 on June 9, 2026. The company described Fable as a Mythos-class model for general use and attributed the different names to their safeguards. It traces Fable's name to the Latin *fabula*, "that which is told," alongside the Greek *mythos*.

Claude Fable 5.1 and Mythos 5.1 followed on September 1, 2026. Fable 5.1 is generally available, while Mythos 5.1 is available through trusted access programs for cybersecurity and life sciences. The Claude API identifier for Fable 5.1 is `claude-fable-5-1`.

## Scope and limits

Fable is a Claude line, not a general technical class of model. Sharing an underlying model with Mythos does not make the deployed products interchangeable. A safeguard can block a request or cause another model to serve it, affecting both capability and the interpretation of evaluation results.

Anthropic acknowledges false positives in its safeguards. The permitted work and fallback behavior depend on the release and policy category. Its documentation recommends observing refusals and the model that actually serves a response; an HTTP success alone does not establish task completion.

## Operational significance

Evaluate the deployed safeguards and fallback path alongside the model. Record which release served each result, whether it completed the requested task, and what additional tool or review work followed. A long-running agent still needs bounded permissions, checkpoints, and independent verification.

Compare cost per accepted task at the chosen effort setting. A high-capability model's published benchmark score does not establish that it is the least expensive option for routine work.

## Distinguish it from nearby terms

- Claude is the parent family; Fable is one named line in it.
- Mythos shares Fable 5.1's underlying model, with different safeguards and restricted access. The deployment distinction is part of their definitions.
- Opus is a separate Claude line. Fable's introduction does not turn Opus releases into aliases of Fable.
- Model routing chooses which model runs. A safeguard-triggered fallback changes the serving path without establishing a new underlying model architecture.

## Check your understanding

Your Fable workflow succeeds after a safeguard sends one request to Opus. Which model and policy details should appear in the evaluation record, and what would a Fable-only success rate leave unexplained?

## Also called

Fable

## Related terms

- [Claude](https://darkfactory.dev/glossary/claude)
- [Claude Mythos](https://darkfactory.dev/glossary/mythos)
- [Claude Opus](https://darkfactory.dev/glossary/opus)
- [Reasoning model](https://darkfactory.dev/glossary/reasoning-model)
- [Agent harness](https://darkfactory.dev/glossary/agent-harness)
- [Model routing](https://darkfactory.dev/glossary/model-routing)
- [Evaluation (eval)](https://darkfactory.dev/glossary/evaluation)
- [Prompt injection](https://darkfactory.dev/glossary/prompt-injection)

## Related factory areas

- [Model selection, routing & budgets](https://darkfactory.dev/factory/model-routing-budgets)
- [Security, privacy & software supply chain](https://darkfactory.dev/factory/security)
- [Verification, evaluation & quality truth](https://darkfactory.dev/factory/verification)

## Evidence and further reading

- [Anthropic: Claude Fable 5 and Claude Mythos 5](https://www.anthropic.com/news/claude-fable-5-mythos-5)
- [Anthropic: Claude Fable 5.1 and Claude Mythos 5.1](https://www.anthropic.com/claude-fable-and-mythos-5-1)
- [Anthropic: Claude models overview](https://platform.claude.com/docs/en/models/overview)
- [Anthropic: Claude refusals and fallback](https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback)