Claude Fable 5 lasted three days. Every alternative is cheaper.
Anthropic shipped its most capable model on June 9 and switched it off on June 12, after a US export-control order it says it disagrees with. If you wired anything to Fable 5 in that window, you need a new model this week. The one upside: Fable 5 was the priciest thing Anthropic has ever sold, so wherever you land, the bill goes down.
Photo by Ash Willson on Unsplash
On June 12, 2026, a US export-control directive pushed Anthropic to disable claude-fable-5 and claude-mythos-5 for every customer worldwide - the order named foreign nationals over a Fable 5 jailbreak, and Anthropic could not filter by nationality in time, so both models went off entirely while the rest of the Claude lineup kept running. The practical problem is narrow but sharp: anyone who adopted the new $10/$50 flagship in its three live days now has a dead model ID. The consolation is that Fable 5 was the priciest thing Anthropic ever sold, so every exit is cheaper - Opus 4.8 halves the rate at $5/$25, and GLM 5.2 ($1.40/$4.40) or Gemini 3.5 Flash ($1.50/$9) take a 20M-in, 4M-out month down by nearly nine tenths.
What actually happened on June 12
Anthropic says it received a government directive at 5:21 PM ET on June 12, citing national-security authorities, and disabled access to both models within hours. The trigger, per Anthropic's own account, was a narrow jailbreak of Fable 5 - which it described as asking the model to read a specific codebase and fix its software flaws. The company called the recall a misunderstanding, said it disagrees that a non-universal jailbreak warrants pulling a model used by hundreds of millions, and that it is working to restore access.
The directive applied to any foreign national, inside or outside the US, including Anthropic's own foreign-national staff. There was no way to cleanly block by nationality on the order's timeline, so the models went dark globally. The Trump administration's position, voiced by adviser David Sacks, is that Anthropic refused to fix the jailbreak before the controls landed; Anthropic disputes that framing. We are reporting the sequence, not refereeing the dispute. What matters for your stack is simpler: the API IDs return nothing right now, and nobody has promised a restore date.
Who this hits, and who barely notices
Mythos 5 was never broadly available. It shipped through Project Glasswing to approved customers and ran without the safety classifiers that Fable 5 carries. Most developers never had a key for it, so its disappearance is a non-event for the general API audience. Fable 5 is the real loss: it was generally available on the Claude API, AWS, Vertex AI, and Microsoft Foundry from June 9, and any team that adopted the new flagship in those three days now has a dead model ID in production.
If you never moved off Opus, you are fine - Opus, Sonnet, and Haiku all kept serving. The pain is concentrated on early adopters who rushed to the top of the lineup the week it landed. That is a small group, but it is exactly the group running the most expensive, highest-stakes workloads, which is why the replacement choice is worth getting right.
Fable 5 sat at the top of the price list
At $10 input and $50 output per million tokens, Fable 5 was the most expensive model Anthropic had ever published - double Opus 4.8 on input and double on output. You paid for a 1M-token context, a 128K output ceiling, and always-on adaptive thinking. With the model gone, the question is not whether a replacement is cheaper. It is how much cheaper you want to go, and what you give up to get there.
| Model | Input / 1M | Output / 1M | Context | Why pick it |
|---|---|---|---|---|
claude-fable-5 (disabled) | $10.00 | $50.00 | 1M | Unavailable since June 12 |
claude-opus-4-8 | $5.00 | $25.00 | 1M | Same vendor, same shape, half price |
gpt-5.5 | $5.00 | $30.00 | 1.05M | Frontier swap if you already run OpenAI |
gemini-3.1-pro | $2.00 | $12.00 | 1M | Cheaper frontier, smaller output ceiling |
gemini-3.5-flash | $1.50 | $9.00 | 1M | Cheapest first-party hosted option |
glm-5.2 | $1.40 | $4.40 | 1M | Open weights, lowest list price |
One caveat on the cheap end. GPT-5.5 charges a 2x input and 1.5x output surcharge above 272K tokens, so if your Fable 5 prompts genuinely used the long context, its effective rate climbs. Gemini 3.1 Pro and 3.5 Flash cap output at 64K against Fable 5's 128K, which matters for long generations. GLM 5.2 has the lowest sticker but runs a peak-hours surcharge on its own API during Beijing afternoons. None of these are dealbreakers - they are the kind of fine print that turns a headline rate into a real invoice.
What the swap does to a monthly bill
Take a flagship-tier workload: 20M input and 4M output tokens a month, the shape of an agentic pipeline that reserves its best model for the hard calls. Here is what each replacement does to that bill against the Fable 5 baseline.
| Model | Input cost | Output cost | Monthly | vs Fable 5 |
|---|---|---|---|---|
| Fable 5 (disabled) | $200.00 | $200.00 | $400.00 | baseline |
| Opus 4.8 | $100.00 | $100.00 | $200.00 | -50% |
| GPT-5.5 | $100.00 | $120.00 | $220.00 | -45% |
| Gemini 3.1 Pro | $40.00 | $48.00 | $88.00 | -78% |
| Gemini 3.5 Flash | $30.00 | $36.00 | $66.00 | -84% |
| GLM 5.2 | $28.00 | $17.60 | $45.60 | -89% |
The spread is wide because Fable 5 sat so high to begin with. Staying inside Anthropic on Opus 4.8 already halves the cost with no integration work beyond the model string. Crossing to GLM 5.2 cuts almost nine tenths, but you are now on a different vendor, a different SDK, and a model with its own benchmark story. Plug your real token counts into the cost calculator before you commit - the right answer changes a lot with your input-to-output ratio.
How we'd choose under deadline
With a dead model ID in production and no restore date, the priority is a replacement you can ship today, not the theoretical best. Opus 4.8 is the obvious first move. It is the same vendor, the same 1M context, the same 128K output, and consumer Claude already falls back to it - so the behavior is the closest match you will find and the integration is a near no-op. The one snag: on Opus 4.7 and later, temperature, top_p, and top_k return a 400 if set to non-default values. Strip them and steer through the prompt.
If this episode has you rethinking single-vendor exposure - a reasonable reaction to watching a flagship vanish on three days' notice - the move is to add a second provider behind a thin routing layer rather than to re-platform in a panic. GPT-5.5 is the cleanest cross-frontier swap on capability. Gemini 3.5 Flash is the value pick if your work tolerates a smaller output window. GLM 5.2 is the budget floor, and its open weights mean you could self-host if you ever face the same kind of access risk again.
The lesson the early Fable 5 adopters paid for: a model being live today is not a guarantee it is live next week, and the reason it goes away may have nothing to do with you. Pinning to one model ID with no fallback is now a real operational risk, not just a portability nicety.
FAQ
Why did Anthropic disable Fable 5 and Mythos 5?
A US export-control directive received June 12, 2026 at 5:21 PM ET, citing national security over a narrow Fable 5 jailbreak. The order targeted foreign nationals; Anthropic could not filter by nationality fast enough and pulled both models globally. All other Claude models stayed online.
How much did Fable 5 cost?
$10 input and $50 output per million tokens, 1M context, 128K max output - the most expensive model Anthropic ever listed. Mythos 5 matched the rate but was a limited Project Glasswing release.
What is the closest replacement?
Claude Opus 4.8 (claude-opus-4-8): same vendor, same 1M context and 128K output, at $5/$25 - half Fable 5's rate. For deeper cuts, GLM 5.2 ($1.40/$4.40) or Gemini 3.5 Flash ($1.50/$9).
Will Fable 5 come back?
Anthropic disagrees with the order and says it is working to restore access, but has given no date. Migrate now - the timeline hinges on a government decision, not a code fix.
Is the swap just a model-string change?
To Opus 4.8, almost - remove temperature, top_p, and top_k first, since they 400 on Opus 4.7 and later. Moving to GPT-5.5, Gemini, or GLM 5.2 is a full integration on a different SDK.
Sources
- Anthropic statement on Fable 5 and Mythos 5 access
- Anthropic docs - introducing Claude Fable 5 and Mythos 5 (specs and pricing)
- CNBC - Anthropic disables Fable 5 and Mythos 5 to comply with government directive
- Al Jazeera - US orders Anthropic to disable AI models for foreign nationals
- VentureBeat - enterprise impact and what to do
- Anthropic models overview and pricing (Opus 4.8)