sakutto
Generative AI· Fable 5

What Is Claude Fable 5? Why It's Offline, Alternatives, and What's Next

ClaudeAnthropicFable 5Export Control

What Claude Fable 5 Is and How It Was Shut Down

Claude Fable 5 Timeline (June 2026)

From release to worldwide shutdown and end of free trial. Source: Anthropic official news.

Jun 9Fable 5 Released Free trial begins for Pro / Max / Team / Enterprise
Jun 12US government issues export control directive Demands suspension of access for all foreign nationals
Same dayAnthropic disables models for all customers Fable 5 and Mythos 5 taken offline worldwide
Jun 22Free trial period ends (model still offline)
Jun 23Day 11 offline. No recovery date announced

Claude Fable 5 went from launch to worldwide unavailability in just three days—an unprecedented event for a commercial AI model. Here is what Fable 5 is, why it was shut down, and where things stand now.

Fable 5 Is Anthropic's Highest-Performing Model

Claude Fable 5 is Anthropic's highest-performing AI model, released on June 9, 2026. Anthropic describes it as "a Mythos-class model made safe for general use." Mythos is the top-tier model line restricted to cybersecurity specialists, and Fable 5 was designed to bring that level of capability to the general public with additional safety layers.

View official source →
"Today we're launching Claude Fable 5: a Mythos-class model that we've made safe for general use." — from the release announcement

Fable 5 achieved the highest scores among frontier models on both coding and analytics benchmarks, becoming the first model to break 90% on Anthropic's core analytics benchmark. This represented a 10-point jump over the previous generation, Opus 4.8—a significant leap rather than an incremental update.

View official source →
"Claude Fable 5 is the first to break 90% on our core analytics benchmark of complex, long-running analytical tasks — a 10-point jump over Opus." — from the release announcement

API pricing was set at $10 per million input tokens and $50 per million output tokens—less than half the price of Mythos Preview, making it comparatively accessible for a frontier model.

View official source →
"$10 per million input tokens and $50 per million output tokens—less than half the price of Claude Mythos Preview" — from the release announcement

Mythos 5, released on the same day, is a version with some safety restrictions removed, provided to cybersecurity specialists through Project Glasswing, a collaboration with the US government. Fable 5 retains Mythos 5's performance with additional safety measures for general use. For a broader overview of the Claude model lineup, see our guide to Claude AI.

From Launch to Worldwide Shutdown in 3 Days

Alongside the release, Anthropic set up a two-week free trial from June 9 to 22 for Pro, Max, Team, and seat-based Enterprise users. Many users began experimenting with the model immediately. Then, on June 12, everything changed.

View official source →
"From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost." — from the release announcement

The US government issued an export control directive, and Anthropic disabled Fable 5 and Mythos 5 for all customers worldwide on the same day. While the directive targeted foreign nationals specifically, Anthropic shut down the models for all users—including US nationals—to ensure compliance.

View official source →
"we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance" — from the official statement

Within three days of launch, the model became unavailable to everyone. A commercial AI model being shut down by a government national security directive on the same day is unprecedented, and the possibility of similar actions affecting other AI companies cannot be ruled out.

What the Export Control Directive Says

The directive cites national security authorities and requires the suspension of access to Fable 5 and Mythos 5 for all foreign nationals, whether inside or outside the United States. Even Anthropic's own foreign national employees are included.

View official source →
"The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." — from the official statement

The directive applies only to Fable 5 and Mythos 5—other Claude models like Opus 4.8 and Sonnet 4.6 are unaffected. The Claude service as a whole is not down; only the two newest models are frozen.

The Jailbreak and Safety Debate Behind the Shutdown

Anthropic's Defense-in-Depth Approach

The layered safety architecture applied to Fable 5. Based on Anthropic's official statement.

1Keep narrow jailbreaks narrow Constrain scope to prevent generalization
2Make universal jailbreaks extremely expensive Raise cost barriers for attackers
3Monitor to detect and shut down successful attacks Catch breaches post-hoc
4External red teaming and bug bounty Over 1,000 hours of testing with no universal jailbreaks found

Behind the shutdown is a fundamental disagreement between the US government and Anthropic over Fable 5's security. The government viewed the risk as severe enough to warrant an immediate recall; Anthropic argues its safety measures make the directive disproportionate.

The Security Vulnerability the Government Flagged

The government's concern centers on a "narrow jailbreak technique" with potential cybersecurity applications. According to Anthropic, the technique involves asking the model to read a specific codebase and fix software flaws.

View official source →
"essentially consists of asking the model to read a specific codebase and fix any software flaws" — from the official statement

Why is fixing vulnerabilities a problem? From an attacker's perspective, finding and fixing vulnerabilities is two sides of the same coin. A model that can identify software weaknesses with high accuracy is a tool that works for both defense and offense. This dual nature is known as the "dual-use problem" and has been debated in AI safety circles for years.

Anthropic characterizes the technique as "narrow," but the government concluded it warranted recalling a commercial model. The disagreement comes down to different thresholds for risk tolerance and how granularly risk should be assessed.

Anthropic's Defense-in-Depth and Pushback

Anthropic has explicitly pushed back against the directive, arguing that recalling a commercial model serving hundreds of millions of users over a narrow jailbreak finding is an overreaction.

View official source →
"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people." — from the official statement

Anthropic points to its "defense in depth" approach: narrow jailbreaks are kept narrow, universal jailbreaks are made extremely expensive to produce, and monitoring catches successful attacks quickly.

View official source →
"We aimed to make jailbreaks either narrow (in the case of non-universal jailbreaks) or very expensive to produce (in the case of universal jailbreaks), and to combine this with thorough monitoring to quickly detect and shut down any successful attacks." — from the official statement

Anthropic's position is that Fable 5's safeguards are "substantially more effective than those of any previously deployed model." Testing involved red teams from the US government, UK AISI, multiple third-party firms, and internal teams. An external bug bounty of over 1,000 hours found no universal jailbreaks in long-running agentic tasks.

View official source →
"ran an external bug bounty that produced no universal jailbreaks in over 1,000 hours of testing" — from the release announcement
View official source →
"Fable's safeguards are substantially more effective than those of any previously deployed model" — from the official statement

Free Trial End and Billing

The Fable 5 free trial period (June 9–22) ended while the model remained offline. Usage credits were supposed to be required starting June 23, but since the model is unavailable, the transition to paid usage has not occurred.

View official source →
"On June 23, we'll remove Fable 5 from those plans. Using it after that will require usage credits." — from the release announcement

Anthropic has not addressed compensation for the downtime. Pro and Max plan subscribers can still use other models (Opus 4.8, Sonnet 4.6), so the plans remain functional. API usage fees for Fable 5 are not being charged since no tokens are being consumed while the model is offline.

Whether Pro plan pricing or rate limits will change if the shutdown continues long-term remains unclear. For now, maintaining your subscription and waiting for recovery is the practical choice, as other models continue to work normally.

Alternatives and Recovery Outlook

Claude Model Availability (as of June 23, 2026)

Only Fable 5 and Mythos 5 are offline. All other models remain available. Source: Anthropic official statement.

Opus 4.8Top-tier general-purpose model for long tasks and advanced analysis
Sonnet 4.6Balanced model for everyday tasks to moderate complexity
HaikuLightweight and fast for short queries and bulk processing
Fable 5Offline due to export control directive
Mythos 5Offline due to export control directive

With Fable 5 unavailable, what should Claude users do? Only Fable 5 and Mythos 5 are affected—Anthropic's other models continue to operate normally. Here is a look at available alternatives, the recovery outlook, and how to prepare for similar disruptions.

Opus 4.8 and Sonnet 4.6 Remain Fully Available

Anthropic has explicitly stated that models other than Fable 5 and Mythos 5 are unaffected. Opus 4.8, Sonnet 4.6, and Haiku are all fully available via both the API and the web interface (claude.ai).

View official source →
"Access to all other Anthropic models will not be affected." — from the official statement

Most tasks that Fable 5 handled can be covered by Opus 4.8. Opus 4.8 was the top-tier general-purpose model before Fable 5's release and handles coding, long-form analysis, and autonomous tasks. Check Anthropic's official documentation for the latest specifications.

That said, Fable 5 represented a 10-point jump over Opus 4.8 on analytics benchmarks. Advanced workflows built specifically for Fable 5 may see reduced accuracy when switching to Opus 4.8.

Recovery Timeline and How to Track Updates

Anthropic has said it is working to restore access as soon as possible, but has not provided a specific timeline. Negotiations with the government are ongoing, and recovery likely depends on multiple factors including enhanced safety measures or changes to the directive's conditions.

View official source →
"We are working to restore access as soon as possible." — from the official statement

As of June 23, the model has been offline for 11 days. A near-term recovery is not realistic. Export control directives are based on government national security assessments and cannot be resolved through technical fixes alone.

The most reliable way to track recovery status is to check Anthropic's official news page directly. Social media and aggregator sites tend to be inaccurate on timing and conditions, especially given that AI regulation involves national policy considerations.

Using Multiple Models to Reduce Shutdown Risk

The Fable 5 shutdown highlighted the risk of depending on a single AI model. Even the most capable model can be taken offline without warning due to regulation or national security decisions.

An effective way to manage this risk is to spread your AI usage across multiple models and providers. For example:

  • Day-to-day coding assistance: Claude Sonnet 4.6 or Claude Opus 4.8
  • Long-form analysis and summarization: Claude Opus 4.8
  • Quick queries and bulk tasks: Claude Haiku
  • Non-Claude alternatives: ChatGPT (GPT-5 series) or Google Gemini as backup

Keeping at least two models capable of handling your core workflows means a sudden shutdown of one won't halt your work. This is the simplest way to ensure that events like this don't disrupt your operations.

The practical benchmark is whether your next workday still runs if your primary model goes offline. For coding, keep prompts that work with both Opus 4.8 and GPT-5. For document summarization, test whether Gemini or Perplexity can serve as a fallback. That level of preparation significantly reduces the impact of shutdowns like this one.

For a comparison of ChatGPT's model lineup, see our guide to ChatGPT and GPT-5. For the full Claude model overview, see our guide to Claude AI.

Free ToolURL to Markdown ConverterConvert any public web page URL to Markdown. Preserves headings, tables, lists, and links — perfect for LLM and RAG preprocessing, research notes, and archiving web articles.Try it now →

Official Sources for Claude Fable 5

This article is based on the following primary sources from Anthropic. Availability and recovery outlook may change, so always check the official sources for the latest information.

Anthropic Official Statement — Fable and Mythos accessView official source →
Anthropic Official Release — Claude Fable 5 and Claude Mythos 5View official source →

FAQ

Q. What is Claude Fable 5?
Claude Fable 5 is an AI model released by Anthropic on June 9, 2026, offering Mythos-class performance with safety measures for general use. It achieved the highest scores among frontier models on coding and analytics benchmarks, but was taken offline worldwide shortly after launch due to a US government directive.
Today we're launching Claude Fable 5: a Mythos-class model that we've made safe for general use. Anthropic Official Release
Q. Why is Claude Fable 5 unavailable?
The US government issued an export control directive citing national security authorities, requiring Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals. Anthropic disabled the models for all customers to ensure compliance.
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States Anthropic Official Statement
Q. When will Claude Fable 5 come back online?
No specific timeline has been announced. Anthropic has stated they are working to restore access as soon as possible, but negotiations with the government are ongoing and no date has been set.
We are working to restore access as soon as possible. Anthropic Official Statement
Q. Are other Claude models still available during the Fable 5 shutdown?
Yes. Opus 4.8, Sonnet 4.6, and Haiku are unaffected and remain fully available. Only Fable 5 and Mythos 5 are subject to the export control directive.
Access to all other Anthropic models will not be affected. Anthropic Official Statement
Q. Does the Fable 5 shutdown affect users outside the US?
Yes. The directive targets all foreign nationals regardless of location, and Anthropic disabled the models for all customers worldwide, including US nationals, to ensure compliance.
suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees Anthropic Official Statement
Q. What was the jailbreak that caused the shutdown?
According to Anthropic, the technique involved asking the model to read a specific codebase and fix software flaws. The government considered this a cybersecurity risk, while Anthropic assessed it as a narrow, limited vulnerability.
essentially consists of asking the model to read a specific codebase and fix any software flaws Anthropic Official Statement
Q. Does Fable 5 have safety problems?
Anthropic maintains that Fable 5's safeguards are substantially more effective than those of any previously deployed model. Over 1,000 hours of external bug bounty testing found no universal jailbreaks.
Fable's safeguards are substantially more effective than those of any previously deployed model Anthropic Official Statement
Q. What happened to the Claude Fable 5 free trial?
The free trial period for Pro, Max, Team, and Enterprise users (June 9–22) ended while the model remained offline. The transition to paid usage on June 23 has not occurred because the model is unavailable.
On June 23, we'll remove Fable 5 from those plans. Using it after that will require usage credits. Anthropic Official Release
Q. How much does Claude Fable 5 cost?
API pricing is $10 per million input tokens and $50 per million output tokens, less than half the price of Claude Mythos Preview. However, the model is currently offline and unavailable for use.
$10 per million input tokens and $50 per million output tokens—less than half the price of Claude Mythos Preview Anthropic Official Release

Related Tools

Related Tool Categories

Articles