Lesson 3 · Claude Mastery
Free
~7 min read
Updated July 19, 2026
Claude Fable 5: the shutdown and the comeback.
Fable 5 had the strangest first month of any AI model: launched June 9, 2026, switched off three days later by a US government export-control order, and brought back on July 1 with a new safety classifier. If you're wondering whether you can use it — you can. Here's the full story and where it now fits.
Update · July 19, 2026: the transition window has closed. Included Fable 5 access on paid plans (up to 50% of weekly usage) was extended twice — from July 7 to July 12, then to a final
July 19, 2026 (11:59 PM PT). Anthropic says Fable 5 isn't permanently leaving subscriptions — it aims to restore standard access when capacity allows. Since July 20 (and past the 50% allowance during the window), access runs on
usage credits at standard API rates ($10/M input, $50/M output). How that works, and how to cap it:
what Claude actually costs.
01 The short version
Claude Fable 5 is available again. After launching June 9, 2026 and being switched off three days later under a US government export-control order, it was reinstated on July 1, 2026 — globally, across Claude.ai, the API, Claude Code, and Cowork. The comeback came with a new cybersecurity safety classifier, and the model sits at premium pricing: $10 per million input tokens / $50 per million output.
In one line
Fable 5 was off for 18 days by government order, and has been back since July 1 — with a new safety layer and a premium price tag. The rest of this lesson is the full story.
02 The timeline
June 9, 2026 — Launch
Anthropic releases Claude Fable 5, the first publicly available model in its top-tier "Mythos" class — positioned above the Opus line.
June 12, 2026 — Government directive
The US Commerce Department, in a letter from Secretary Howard Lutnick, directs Anthropic to suspend access for all foreign nationals worldwide, following reports of a jailbreak that could bypass the models' safety guardrails. Because the order reaches foreign nationals everywhere, Anthropic disables Fable 5 and Mythos 5 globally, for everyone.
June 30, 2026 — Controls lifted
The Commerce Department lifts the export controls after review.
July 1, 2026 — Reinstated
Anthropic redeploys Fable 5 worldwide — Claude.ai, API, Claude Code, and Cowork — alongside a new cybersecurity classifier addressing the jailbreak that triggered the order.
03 Why it happened — and what changed on the way back
The original order followed reports of a jailbreak — a technique for bypassing the model's safety guardrails that, in the government's view, could turn a frontier model into an unrestricted cyber tool. Anthropic complied with the directive while publicly disagreeing that a narrow technique justified recalling a deployed model.
The version that returned isn't simply switched back on. Anthropic trained a new safety classifier that blocks the specific technique in the overwhelming majority of cases — and when a request trips it, the system routes the answer to Opus 4.8 instead of refusing outright. In practice: a small number of borderline prompts get answered by a different (still excellent) model rather than Fable 5 itself.
If a Fable 5 response occasionally arrives labeled as Opus 4.8, that's the classifier doing its job — not a bug, and not your account being restricted.
04 Should you actually use it?
The Claude lineup changed twice the same week Fable 5 returned — Sonnet 5 became the new default on June 30. So the honest ladder now:
- Sonnet 5 — the default for Free and Pro; near-flagship performance and the right starting point for nearly everything.
- Opus 4.8 — the proven heavyweight for genuinely hard reasoning, coding, and writing.
- Fable 5 — the Mythos-class frontier model at premium API pricing ($10/$50 per million tokens). Reach for it when a task earns it: the longest autonomous work, the hardest problems, where the frontier margin matters.
Start on Sonnet 5, escalate deliberately. For the cost math, see what Claude actually costs.
Frequently asked
Fable 5 — your questions, answered
Is Claude Fable 5 available again?
Yes. Fable 5 was reinstated on July 1, 2026 after the US Commerce Department lifted its export controls on June 30. It's live globally across Claude.ai, the API, Claude Code, and Cowork.
Why was Fable 5 suspended?
A US Commerce Department directive (a letter from Secretary Howard Lutnick) ordered Anthropic to suspend access for all foreign nationals worldwide on June 12, 2026, after reports of a jailbreak that could bypass the model's safety guardrails. Anthropic disabled the model globally to comply, while publicly disagreeing that the technique justified a recall.
What changed when Fable 5 came back?
Anthropic shipped a new cybersecurity safety classifier that blocks the jailbreak technique in the vast majority of cases. When a request trips the classifier, the answer is routed to Opus 4.8 instead of Fable 5 — so borderline prompts still get answered, just by a different model.
What does Fable 5 cost?
API pricing is $10 per million input tokens and $50 per million output — roughly 4–5× Sonnet 5's standard rate. It's the premium tier of the lineup.
Should I use Fable 5, Opus 4.8, or Sonnet 5?
Start with Sonnet 5 — the new default (June 30) with near-flagship performance. Escalate to Opus 4.8 for genuinely hard tasks, and to Fable 5 when frontier capability is worth frontier pricing.
Could Fable 5 be pulled again?
The June suspension shows frontier models can be subject to government action, so it's not impossible — but the export controls were lifted after review and the model returned with added safeguards. Treat it as a normal product with an unusual origin story, and keep critical workflows portable across models as good practice.