The short answer (July 2026): Claude Fable 5 is Anthropic's most capable generally available model, released June 9, 2026 as the first of a new "Mythos-class" tier that sits above the Opus line. It costs $10 per million input tokens and $50 per million output tokens on the API (model ID claude-fable-5), ships with a 1 million token context window, and is available on Claude Pro, Max, Team, and Enterprise plans and in Claude Code. It launched, was suspended under US export controls on June 12 after a safeguard bypass was discovered, and was restored globally on July 1. Yes, it is available right now.
Last verified: July 10, 2026, against Anthropic's announcement, model docs, and pricing pages.
What Fable 5 actually is
Anthropic released two models on June 9 with one underlying brain. In the company's announcement, Claude Fable 5 is "a Mythos-class model made safe for general use," and Claude Mythos 5 is "the same underlying model as Fable 5, but with the safeguards lifted in some areas." Mythos-class means "a tier of Claude models that sit above our Opus class in capability." Both succeed the Claude Mythos Preview that Anthropic showed in April 2026.
The practical translation: Fable 5 is the one you can use. Mythos 5 is invitation-only, restricted to approved organizations in Anthropic's Project Glasswing defensive-cybersecurity program, with no self-serve signup.
On capability, Anthropic claims "state-of-the-art on nearly all tested benchmarks," with standout results in software engineering, knowledge work, vision, and scientific research, including the top score on Cognition's FrontierCode evaluation. Those are the vendor's own numbers; treat them as a strong signal, not gospel. The credible third-party endorsement in the launch material is Stripe's, saying Fable 5 "compressed months of engineering into days."
The June suspension, explained
You may have found this page because Fable 5 vanished two weeks ago. What happened, per Anthropic's own account: on June 12, Amazon researchers reported a method that bypassed Fable 5's cybersecurity safeguards, and the US government applied export controls to Anthropic's newest models. Unable to verify user nationality in real time, Anthropic suspended Fable 5 and Mythos 5 for everyone.
The controls were lifted June 30, and access was restored globally on July 1 across the Claude Platform, claude.ai, Claude Code, and Claude Cowork, with cloud-marketplace reinstatement on AWS, Google Cloud, and Microsoft Foundry following. The fix: a retrained safety classifier that Anthropic says blocks the reported technique "in over 99% of cases," with blocked requests falling back to Claude Opus 4.8.
We covered the story as it unfolded; the Anthropic news feed has the running record.
How to get access
- claude.ai plans: Fable 5 is on Pro, Max, Team, and seat-based Enterprise. It was included at no extra cost from launch through June 22; since then it draws on usage credits on those plans, per the announcement.
- Claude Code: available since the July 1 restoration. Note one docs detail: thinking cannot be disabled on Fable 5, it always runs extended thinking.
- API: model ID
claude-fable-5, generally available on the Claude API, Amazon Bedrock (anthropic.claude-fable-5), Google Cloud Vertex AI, and Microsoft Foundry. - Mythos 5: you don't, unless your organization is in Project Glasswing. Anthropic points prospective partners to their account teams.
Pricing and specs
| Model | Input / Output per MTok | Positioning |
|---|---|---|
| Fable 5 | $10 / $50 | Highest available capability |
| Opus 4.8 | $5 / $25 | Default for complex agentic coding and enterprise work |
| Sonnet 5 | $3 / $15 (intro $2 / $10 through Aug 31) | Everyday workhorse |
| Haiku 4.5 | $1 / $5 | Speed and scale |
The specs that matter, from the model docs:
- Context window: 1M tokens by default; up to 128K output tokens per request.
- Knowledge cutoff: January 2026 (reliable), the most current of any Claude model.
- Adaptive thinking, always on. There is no non-thinking mode; you steer depth with the
effortparameter instead. Raw chain of thought is never returned; you can request a summarized version. - Tokenizer: the one introduced with Opus 4.7, which counts roughly 30% more tokens than older Claude models for the same text. Budget accordingly when comparing costs to pre-4.7 models.
The safety classifiers, and when you'll notice them
This is the genuinely novel part of Fable 5's design and worth understanding before you build on it. Fable 5 ships with separate classifier models watching for misuse in three areas: cybersecurity, biology and chemistry, and large-scale distillation attempts. When a request trips one, Fable 5 doesn't answer it; per Anthropic, "requests flagged as cybersecurity will instead receive a response from our next-most-capable model, Claude Opus 4.8," and most biology and chemistry requests fall back the same way.
Anthropic says the classifiers "trigger, on average, in less than 5% of sessions." For API builders, the mechanics matter: a declined request returns stop_reason: "refusal" as an HTTP 200, not an error, and you are not billed when it refuses before producing output. Anthropic offers a server-side fallbacks parameter (in beta) and SDK middleware to route refusals to another model automatically. If your workload lives in security research or the life sciences, test against these classifiers before committing; everyone else will rarely meet them.
Should you use it?
Use Fable 5 when the task is genuinely hard: long multi-step agentic work, research-grade analysis, complex codebases, vision-heavy workloads, or anything where a weaker model's failure costs more than the price gap. At double Opus 4.8's price and five times Sonnet's, it is poor economics as a default model for routine work, and Anthropic's own docs position Opus 4.8 as the everyday enterprise default. The 1M context window is the underrated feature: entire codebases or document sets in one request, no retrieval scaffolding.
For how Fable 5 changes the assistant you use day to day, see our Claude guide; for the model landscape it competes in, ChatGPT vs Claude vs Gemini.
FAQ
Is Claude Fable 5 available right now?
Yes. Restored globally on July 1, 2026 on claude.ai, the API, Claude Code, and Cowork, after the June 12 export-control suspension.
What's the difference between Fable 5 and Mythos 5?
Same model, different safeguards. Fable 5 carries misuse classifiers for cybersecurity, bio/chem, and distillation. Mythos 5 lifts the cyber safeguards for vetted defensive-security partners only.
Is Fable 5 better than Opus 4.8?
More capable, per Anthropic's benchmarks, and twice the price. Opus 4.8 remains their recommended default for most serious work; Fable 5 is for the tasks that need the ceiling.
Why is it called Fable?
Anthropic hasn't published the naming rationale. What's documented is the tier logic: Mythos is the class name, Fable 5 is its generally available member.
Model launches, pricing shifts, and the occasional export-control drama: this is exactly what we track. This page is re-verified whenever Fable 5's status changes. Get AI Weekly free, 3 issues a week, read by 40,000+ practitioners.