Claude Fable 5 and 5.1: Suspension, Return and What Changed
June 9, 2026
Update, October 2026. Claude Fable 5 is back, and it has a successor. Anthropic suspended Fable 5 for all users on June 12, 2026, after a US Commerce Department export control directive. The controls were lifted on June 30, and Anthropic restored access globally on July 1. On September 1, Anthropic released Claude Fable 5.1 at the same base price. On September 22, it released Claude Opus 5.5, which Anthropic says performs at the level of Fable 5.1 on most work at a lower price. I have rewritten this post to reflect where things stand today.
Anthropic released Claude Fable 5 on June 9, 2026, as the most capable model it had made publicly available. Three days later it went offline for everyone. It came back on July 1, and Fable 5.1 replaced it as the newest version on September 1. This post covers what Fable is good at, what happened during the suspension, what changed in 5.1, and which Claude model I would use now.
What is Fable 5, in one paragraph
Anthropic built a model called Mythos 5 that they consider too capable to release openly, keeping it restricted to vetted security and biomedical researchers. Fable 5 is the same model with a safety layer on top. Requests touching dangerous territory (think bioweapons, cyberattack tooling) get quietly rerouted to the older Opus 4.8 model. Anthropic says this happens in fewer than 5% of sessions. For consulting, finance, sales, and marketing work, you would essentially never hit it. The export-control directive that caused the June 12 suspension was triggered specifically by the model's ability to find software vulnerabilities in code. Anthropic has publicly noted that the same capability exists in other deployed models.
What it is best at
The pattern across every early benchmark and test: Fable 5 pulls ahead on long, complex, multi-step work. On short, simple prompts, the difference from Opus 4.8 is hard to feel.
Where the gains were clear at launch:
Financial analysis. Fable 5 scored highest of any model on Hebbia's Finance Benchmark, which tests reasoning at senior-analyst level: reading documents, interpreting charts and tables, solving problems across them. Trading firm IMC reported it passed their trading-analysis evaluations almost across the board.
Data analysis. First model to break 90% on Hex's analytics benchmark, a 10-point jump over Opus 4.8 on long-running analytical tasks that need judgment, not just computation.
Large document sets. It holds focus across millions of tokens and writes its own working notes as it goes. If your work involves a 40-page report, six competitor documents, and a regulatory framework all at once, this is the upgrade you will feel.
Reading visuals. A meaningful jump in chart, table, and screenshot interpretation. In one test it rebuilt a working web app from screenshots alone. Practical version: it reads the graphs in your client's PDF correctly more often.
Contract review. Lawyers in blind tests found its contract redlines matched or beat specialized legal AI tools. It is already integrated inside Harvey, the legal AI platform.
Long-running tasks. Anthropic markets "days-long" autonomous work. One caveat the launch coverage often skipped: that applies to automated pipelines built on the API, not to chatting in the Claude app. In normal use, what you get is a model that stays coherent much deeper into a long working session before it starts losing the thread.
What it is not better at
Quick drafts and short prompts. Early reviewers at Every.to found the model's slower, deeper thinking "works against the rapid-iteration rhythm most writing depends on". If your main use is drafting emails and punching up copy, a faster and cheaper model such as Opus 5.5 or Sonnet 5.5 does the job.
Marketing and sales writing specifically. No independent test of proposal writing, marketing briefs, or sales decks surfaced before the suspension. The finance and analytics gains were confirmed. The writing gains were not.
Finding a use for it at all. The same Every.to review noted that heavy AI users found it paradigm-shifting on their hardest tasks, while lighter users "struggled to find something to use it for." If you have not yet built AI into a real workflow, a stronger model does not change that. The bottleneck is the workflow, not the model.
A short history of the suspension
People still ask me what happened, so here is the timeline from Anthropic's own account:
- June 9: Anthropic released Fable 5 and Mythos 5.
- June 12: A US export control directive required Anthropic to restrict access for foreign nationals. Anthropic could not verify nationality in real time, so it suspended both models for all users.
- June 26: The US government approved restored Mythos 5 access for a set of US organizations.
- June 30: The export controls on both models were lifted.
- July 1: Anthropic restored Fable 5 for users globally on Claude.ai, Claude Code, Claude Cowork and the Claude API.
In June I expected a US-first return with EU users waiting longest. That is not what happened. Access came back for users globally on the same day.
What changed in Fable 5.1
Anthropic released Claude Fable 5.1 on September 1, 2026. Here is what changed, based on Anthropic's release notes for developers and its announcement.
Same base price, cheaper cache reads. It still costs $10 per million input tokens and $50 per million output tokens. Cache reads dropped from $1 to $0.25 per million tokens. Anthropic estimates this saves about 25% on typical workloads and up to about 45% on highly agentic work.
Better at the work it was already good at. Anthropic lists gains in long coding sessions, work with documents, spreadsheets and slides, multistep research, reading dense charts and tables in PDFs, long-context work and computer use.
Fewer false alarms from the safety filters. Anthropic says its cyber safeguards block 60% fewer false positives. Its biology safeguards fire 85% less often on harmless elementary biology and medical questions than at Fable 5's launch.
Same core specs. A 1M token context window, 128K max output and a June 2026 knowledge cutoff.
Some changes for developers. If you call Fable through the API, three changes can break existing code. Forced tool use now returns an error, and thinking blocks are tied to the model and conversation that produced them. Read the migration guide before you switch the model ID.
What your Claude plan includes. According to Anthropic's help center, Max, premium Team seats and premium Enterprise seats include Fable 5 and 5.1 up to 50% of the weekly usage limit. Pro and standard Team seats pay for Fable with usage credits. Standard Enterprise seats need usage credits enabled by the organization. The Free plan has no access.
The data-residency picture for EU users
Here is what Anthropic's docs say today.
Fable requires 30-day data retention. Anthropic classes Fable 5 and Fable 5.1 as Covered Models. On the Claude API they require 30-day data retention and are not available under zero data retention unless Anthropic expressly authorizes it.
The direct Claude API has no EU option. Its data residency controls offer US or global inference only, and US is the only option for data stored at rest.
AWS Bedrock has EU regions, but not for Fable 5.1 yet. Fable 5.1 on Bedrock is open to all customers. Bedrock offers EU regions, but for Fable 5.1 regional endpoints are currently in us-east-1 only. Elsewhere it runs on the global endpoint, which routes across all available regions.
Google Cloud offers an EU multi-region endpoint. Fable 5.1 is listed on Google Cloud's Agent Platform, formerly Vertex AI. Google offers an EU multi-region endpoint for newer Claude models at a 10% premium. Model availability varies by region, so confirm in Model Garden that Fable 5.1 is offered on the EU endpoint before you rely on it.
An earlier version of this post mentioned EU data storage on Claude Enterprise. I could not confirm that in Anthropic's current docs, so ask your account contact if it matters to you. If EU-only processing is a hard requirement, check with your AWS or Google account team before you route client data through Fable.
Which Claude model to use now
Start with Opus 5.5 for most work. Anthropic's own docs say to start with Opus 5.5 for most workloads. Its release notes say Opus 5.5 performs at the level of Fable 5.1 on most work. It costs $4 per million input tokens and $20 per million output tokens, against $10 and $50 for Fable 5.1.
Move to Fable 5.1 for your hardest work. Anthropic positions it for demanding reasoning and long-running agent work, or for cases where Opus 5.5 at higher effort still falls short. For me, that means long financial analysis across many documents, large contract sets and multi-source research.
Use Sonnet 5.5 or Haiku 5.5 for fast, cheap tasks. Sonnet 5.5 costs $2 / $10 per million tokens and Haiku 5.5 starts at $0.10 / $0.50. For quick drafts and routine tasks, they are the faster options.
Build your workflows to be model-agnostic. The suspension showed how fragile a workflow tied to a single model ID can be. Fable 5 was gone for almost three weeks with no warning. If you are setting up templates, automations, or client-facing AI pipelines, build them so you can swap models without rebuilding. That is true for Anthropic models, and for any other provider.
Test your hardest task on both. If you lined up a hard task while Fable was down, run it on Opus 5.5 and Fable 5.1 side by side. Pay for Fable only where you can see the difference.
Check anthropic.com/news and the Claude release notes for updates.
Frequently Asked Questions
Is Claude Fable 5 available right now? Yes. Anthropic restored access to Fable 5 globally on July 1, 2026, on Claude.ai, Claude Code, Claude Cowork and the Claude API. Fable 5.1 has been the newest version since September 1. On Claude plans, Max, premium Team seats and premium Enterprise seats include Fable models up to 50% of the weekly usage limit. Pro and standard Team seats pay for Fable with usage credits. The Free plan has no access.
Why did the US Commerce Department suspend Fable 5? On June 12, 2026, a US export control directive required Anthropic to restrict Fable 5 and Mythos 5 for foreign nationals. Anthropic could not verify nationality in real time, so it suspended both models for all users. The export controls were lifted on June 30, and Anthropic restored access for users globally on July 1.
What is the difference between Fable 5 and Fable 5.1? Fable 5.1 keeps the same $10 input and $50 output price per million tokens, the 1M token context window and the June 2026 knowledge cutoff. Cache reads dropped from $1 to $0.25 per million tokens. Anthropic reports gains in long coding sessions, document and spreadsheet work, research, chart reading and computer use. Its safety filters also flag far fewer harmless requests. Developers should check three breaking API changes before switching.
Should I use Opus 5.5 or Fable 5.1? Start with Opus 5.5 for most work. Anthropic's docs recommend it as the starting point for most workloads, and its release notes say Opus 5.5 performs at the level of Fable 5.1 on most work. It costs $4 / $20 per million tokens, against $10 / $50 for Fable 5.1. Move to Fable 5.1 for demanding reasoning and long-running agent work, or when Opus 5.5 at higher effort still falls short.
What is Fable best at? Long, complex, multi-step analytical work. At launch, Fable 5 set records on Hebbia's Finance Benchmark and was the first model to break 90% on Hex's analytics benchmark. Fable 5.1 adds stronger work with documents, spreadsheets and slides, multistep research, and reading dense charts and tables in PDFs. For short, quick drafts, a faster and cheaper model does the job.
Related articles
Where would AI help your team?
Tell us about your team and the work you want to change. We'll help you pick the right next step.