AI Models & Assistants / Knowledge library
Claude Fable
claude models for extended coding and knowledge work
Why learn Claude Fable?
A candidate for tasks involving sustained reasoning and multiple steps. Evaluate the exact release with its configured tools and access conditions. Claude is the parent family; Cursor documents it as a model option; GPT-6 Astra is an evaluation alternative.
The problem it solves
Targets complex work that requires continuity across reasoning, tools, and verification steps.
Where it came from
Anthropic introduced Claude Fable 5 alongside Mythos 5 in 2026. As of this research snapshot, the Fable product page describes Fable 5.1 for long-running coding, agent, and knowledge-work tasks. This entry tracks the named Fable line within Claude.
Build your understanding
Work through these exercises in order. Try the task before opening a reference.
- Distinguish the Claude family from a Fable release.
- Define acceptance criteria for a multi-step task.
- Evaluate a code change with hidden regression cases.
- Check a document-based answer against tables and figures in the source.
- Measure completed tasks, intervention count, and unsupported claims over repeated runs.
Documentation and sources
- Claude Fable product details (www.anthropic.com)
- Fable 5 introduction (www.anthropic.com)
Catalog research notes
- Checked September 9, 2026: Anthropic's Fable page describes Fable 5.1. Access, retention, and routing conditions are documented there; benchmark claims are vendor-reported.