Claude Fable 5: What It Is, Why It Vanished, and What Its Return Means

Anthropic launched Claude Fable 5 on June 9, 2026 and described it as a Mythos-class model made safe for general use. It arrived alongside Claude Mythos 5, which Anthropic described as the same underlying model with the safeguards lifted in some areas and restricted to approved partners.

The line has been updated since. The current version is Claude Fable 5.2, released on November 9, 2026. Anthropic's documentation recommends it for demanding reasoning and long-horizon agentic work: agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet or deck. For most other workloads, Anthropic says to start with Claude Opus 5.5.

Fable 5.2 has a 1 million token context window and can write up to 128,000 tokens in one response. On the API its model id is claude-fable-5-2. It is a paid model and is not part of the Free plan. For what to do with it, see our Claude Fable 5 use cases and example prompts.

• June 9, 2026, launch. Anthropic introduces Claude Fable 5 and Claude Mythos 5, at $10 input and $50 output per million tokens.
• June 12, 2026, offline. Three days later, the U.S. government issues an export-control directive suspending access to both models by any foreign national. Anthropic says it must disable Fable 5 and Mythos 5 for all customers to comply. Its other models are not affected.
• Mid to late June, the gap. Users hit “Fable 5 currently unavailable” and “invalid or inaccessible model” errors.
• June 30, 2026, restrictions lifted. Commerce Secretary Howard Lutnick tells Anthropic the licensing requirement no longer applies.
• July 1, 2026, restored. Anthropic restores access to Fable 5.
• November 9, 2026, Fable 5.2. Anthropic releases Claude Fable 5.2, the current version, at a lower price of $8 input and $40 output per million tokens.

So to answer the most-searched question directly: yes, Fable is available. The June outage was a government action that has since been reversed, not a bug or a quiet retirement.

The pause was about export controls, not model quality. The directive, issued under national security authorities, suspended access to Fable 5 and Mythos 5 by any foreign national, inside or outside the United States. Anthropic said the practical effect was that it had to switch both models off for everyone.

In its public statement, Anthropic said its understanding was that the government believed it had become aware of a method of bypassing, or “jailbreaking,” Fable 5. That is why “fable 5 jailbreak” searches surged. Anthropic also said it believed the directive was a misunderstanding, and noted it had said at launch that perfect jailbreak resistance is not currently possible for any model provider.

Fable 5.2 costs twice as much per token as Opus 5.5. For heavy automated workloads that difference adds up, which is why Anthropic's own guidance is to start on Opus 5.5 and move to Fable 5.2 when the work is hard enough to need it. Claude Sonnet 5.5 is cheaper again at $2 and $10. See our Claude pricing guide for the full lineup and Claude Opus vs Sonnet for the choice below Fable.

We do not publish benchmark scores here. The launch-era comparisons were against Claude Opus 4.8, which is no longer the current Opus model, and Anthropic's documentation for Fable 5.2 does not list benchmark figures.

Fable and Mythos are the same class of model with different safeguards. At launch, Anthropic said Mythos 5 was the same underlying model as Fable 5 with the safeguards lifted in some areas, and that the safeguards are why the two have different names. Today, Claude Mythos 5.2 shares its specifications and pricing with Claude Fable 5.2, and Anthropic offers it only to organizations verified through its verification programs, such as its Cyber Verification Program.

If you are choosing a model to build with, Fable is the one you can get. Opus 5.5 is the cheaper option below it.

When Fable 5 launched, Anthropic said that requests its classifiers flagged as related to cybersecurity, biology and chemistry, or distillation were automatically handled by Claude Opus 4.8 instead, and that more than 95% of Fable sessions involved no fallback at all. Anthropic's current documentation for Fable 5.2 does not describe the routing in detail, so treat that as the launch design, not a specification of today's behavior.

Anthropic also requires 30-day data retention on Fable. It has said this lets it research and mitigate jailbreaks. For normal work, Fable behaves like any other Claude model, with extra guardrails on a narrow set of sensitive topics.

• Claude apps: Fable is included on Max plans and Team premium seats for up to half of your weekly limit, and on Enterprise plans. Pro subscribers can use it by paying with usage credits. It is not on the Free plan.
• Claude API: call the claude-fable-5-2 model id.
• Cloud platforms: Fable 5.2 is available on Amazon Bedrock, Google Cloud and Microsoft Foundry.
• Claude Code: update to the latest version, then choose Fable as your model for heavy coding sessions.

If you see “Fable 5 currently unavailable” or “invalid or inaccessible model,” update your client, refresh the model list, and check that your plan or API account includes Fable and that you are calling the current model id. For more on working with Claude, see our guides on Claude prompts and using Claude for coding.

For most people: default to Opus 5.5, and move up to Fable 5.2 for the hardest jobs. Opus 5.5 is half the price and has the same context window and output limit. You are unlikely to notice a difference on a quick email or a simple script.

Fable is worth trying on long agent sessions, large refactors and multistep research, where a stronger model can mean fewer wrong turns and less correction. Test it on your own work before committing, because the right answer depends on the task. If you are comparing providers, see ChatGPT vs Claude and Claude vs Gemini.