Fable 5 is back — and the outage is the lesson

The US government lifted the export controls on Claude Fable 5 and Mythos 5 on June 30, and Anthropic restored Fable 5 globally on July 1 — Claude Platform API, Claude.ai, Claude Code, and Cowork — at its original $10/$50 per 1M pricing with the full 1M context window. That closes the loop we opened on June 9: the most capable generally available model on the market spent nineteen days switched off by a government directive, and no amount of subscription money could turn it back on. It now enters our cost calculator for the first time, because "available" finally means available.

Verdict: The export-control suspension is over — Fable 5 is back for everyone at $10/$50, and the three-week outage remains the best argument for local weights

The take

The facts, verified against anthropic.com/news/redeploying-fable-5 (June 30, with the July 1 "access restored" update) and the platform.claude.com pricing page: export controls lifted June 30; Fable 5 available to users globally from July 1; API pricing $10 input / $50 output per 1M (batch $5/$25), 1M context at standard pricing, on the newer tokenizer that produces ~30% more tokens for the same text. Subscription access has a wrinkle: Pro/Max/Team and premium Enterprise plans included Fable 5 within 50% of weekly limits only through July 7 — after that it is usage-credits only on subscriptions. Cloud-platform re-enablement (Bedrock, Vertex, Foundry) was announced as "as quickly as possible" and is phased. Mythos 5 access was separately restored for a set of US organizations following June 26 government approval.

Why it matters here: we held Fable 5 out of the cost calculator through two sweeps — first because you cannot budget around a suspended model, then because the June "restoration" stories turned out to be a plan-billing change plus a narrow institutional deal. Both holds were correct, and the standard we set ("add only on confirmed general restoration at $10/$50") is now met on Anthropic's own pages. At a $30 blended average per 1M it is by far the most expensive row in the calculator — roughly double Opus 4.8 — before the ~30% tokenizer overhead.

Our call: added to the cost calculator; no local-pick change, as ever, for a hosted model. The editorial point stands unchanged from June 9 and is stronger for having resolved: for nineteen days, teams whose workflows depended on the best hosted model had no recourse — while every open-weight model on this site kept running on hardware its owners control. That asymmetry is the thesis of this site. Fable 5 back online is genuinely good news; local weights never needed the news.

Where this fits

Models: GLM-5.1 · Kimi K2.6 · Qwen 3.6-35B-A3B · DeepSeek V4-Pro

Hardware: NVIDIA RTX 5090 · Mac Studio M3 Ultra 96 GB · NVIDIA DGX Spark

Sources

Next step

Try this in the planner→