On June 9, 2026, Anthropic did something it had spent months saying was too risky: it gave the public access to a "Mythos-class" model — the most powerful tier of AI it has ever built. The release, called Claude Fable 5, landed just days after the company publicly warned that frontier AI is becoming dangerously capable and urged the industry to slow down.
That contradiction — racing ahead while waving a caution flag — is exactly what makes this launch worth understanding. Fable 5 is not just another incremental model update. It is the safety-wrapped public version of Mythos, a model so capable at finding software vulnerabilities that Anthropic previously refused to release it to anyone outside a small circle of vetted partners. Here is the full story, explained plainly.
The Contradiction at the Heart of the Launch
To appreciate why Fable 5 is a big deal, you need the timeline. In early June 2026, Anthropic publicly argued that AI labs should consider pausing, warning that humanity risks losing control of increasingly autonomous systems. Then, on June 9, it released its most powerful public model to date.
Anthropic's framing is that Fable 5 resolves the tension rather than ignoring it: instead of keeping all that capability locked away, the company built heavy safety guardrails around it so the public can use the productive parts while the dangerous parts stay blocked. Critics see a company caught between its safety mission and the commercial pressure to ship. Both readings can be true at once — and that is what makes Fable 5 a defining moment for the AI industry in 2026.
What Is Mythos — and Why It Was Locked Away
Mythos is Anthropic's frontier model — a capability tier that sits above the Claude Opus models most people already know. What set it apart, and what made Anthropic nervous, was its skill at cybersecurity.
According to reporting around the model, Mythos can identify and exploit zero-day vulnerabilities — previously unknown software flaws — at a level that surpasses all but the most skilled human hackers. It has reportedly already found thousands of high-severity vulnerabilities, including flaws in every major operating system and web browser.
That kind of capability is a double-edged sword. In the hands of defenders, it can harden the world's software. In the wrong hands, it is a weapon. So Anthropic restricted Mythos to a small group of vetted partners under a program called Project Glasswing, and it has reportedly been involved in government cyber work. Releasing it to the public unmodified was considered too dangerous.
Fable 5 is Anthropic's answer: take that same powerful model, wrap it in conservative safety controls, and let everyone use it for everything except the most sensitive domains.
Fable 5 vs Mythos 5: The Key Difference
This is the part that confuses people, so let's make it simple. Fable 5 and Mythos 5 are the same underlying model. The difference is the guardrails.
| Aspect | Claude Fable 5 | Claude Mythos 5 |
|---|---|---|
| Underlying model | Mythos-class | Mythos-class (identical) |
| Who can use it | The general public | Vetted cyberdefenders & infrastructure providers only |
| Safety guardrails | Active (cyber, bio/chem, distillation) | Some guardrails lifted for approved use |
| Sensitive queries | Rerouted to safer Claude Opus 4.8 | Answered directly in approved domains |
| Access program | Open via API, apps, and cloud | Project Glasswing / trusted access |
In short: Fable is Mythos with a safety harness. If you are a normal developer, researcher, or business, Fable 5 is the version you can use today.
What Fable 5 Can Actually Do
Anthropic describes Fable 5 as state-of-the-art on nearly all tested benchmarks, with one defining trait: it can work autonomously for longer than any previous Claude model. That ability to stick with a long, multi-step task without losing the thread is what unlocks the most impressive results.
Software Engineering
This is where the headline numbers come from. Payments company Stripe reported that Fable 5 "compressed months of engineering into days," completing a migration across a 50-million-line codebase in a single day — work that would have taken a team more than two months by hand. Fable 5 also posted the highest score among frontier models on Cognition's FrontierCode evaluation, and it is more token-efficient than earlier Claude models.
Vision
Fable 5 is the new state-of-the-art for vision tasks. It can extract precise numbers from scientific figures, rebuild a web app's source code from a screenshot alone, and it even completed the game Pokémon FireRed using only vision with minimal scaffolding.
Knowledge Work and Finance
On Hebbia's finance benchmark for senior-level reasoning, Fable 5 was the top performer. Analytics platform Hex reported the model achieved the first 90% score on its core benchmark of complex, long-running analytical tasks, and trading firm IMC said it "aced their trading-analysis evaluations nearly across the board."
Science and Life Sciences
Perhaps the most striking results came from research. Anthropic's internal protein-design experts said Fable 5 accelerated drug-design work roughly tenfold. In blinded comparisons, scientists preferred the model's novel molecular-biology hypotheses about 80% of the time over previous Opus-class models, and it conducted autonomous genomics research matching or exceeding published results.
Memory and Long Tasks
Fable 5 can operate across millions of tokens of context in extended tasks. In one playful test — playing the game Slay the Spire with a persistent file-based memory — it improved three times faster than Claude Opus 4.8.
The Benchmarks and Real-World Results
Here is a snapshot of the concrete results partners and Anthropic reported, gathered in one place.
| Area | Result | Source |
|---|---|---|
| Code migration | 50M-line codebase migrated in 1 day (vs ~2 months) | Stripe |
| Coding benchmark | Highest score among frontier models | Cognition FrontierCode |
| Analytics | First 90% on a complex long-running benchmark | Hex |
| Finance reasoning | Top performer for senior-level tasks | Hebbia |
| Drug design | ~10x faster protein/drug design workflows | Anthropic internal |
| Scientific hypotheses | Preferred ~80% of the time vs Opus-class | Blinded comparison |
| Long-task learning | 3x faster improvement than Opus 4.8 | Slay the Spire test |
The pattern across all of these is consistent: the harder and longer the task, the bigger Fable 5's advantage. It is built for the kind of multi-hour, autonomous work that previous models could not sustain.
The Safety Guardrails Explained
Because Fable 5 is so capable, Anthropic shipped it with the most aggressive safety system it has built. The model uses three main safety classifiers, and when a request trips one, the answer is handled by the safer Claude Opus 4.8 instead of Fable 5.
| Guardrail | What It Blocks | Why |
|---|---|---|
| Cybersecurity | Exploitation and offensive cyber tasks | Mythos can find and exploit real vulnerabilities |
| Biology & Chemistry | Most bio/chem requests, e.g. gene-therapy design | Prevents misuse for bioweapons or harmful agents |
| Distillation | Attempts to extract the model to train a rival | Protects against cloning the model's capabilities |
How Conservative Are These Guardrails?
Very. Anthropic tuned them to err on the side of caution, meaning they will sometimes catch harmless requests. Even so, the company says the safeguards trigger in fewer than 5% of sessions — so more than 95% of the time, you are talking to the full-power Fable 5 with no fallback.
The safety testing was extensive. Anthropic ran an external bug bounty exceeding 1,000 hours that produced no universal jailbreaks, and external partners called the model the most robust they had tested. It also complied with zero harmful single-turn requests across 30 different public jailbreak techniques.
The Privacy Trade-Off
There is one notable catch worth knowing: Anthropic enforces a mandatory 30-day data-retention policy on all Mythos-class traffic — even for enterprises that previously had zero-retention agreements. The data is used only for safety monitoring, not training, and is automatically deleted after 30 days in nearly all cases. Still, it is a meaningful exception to Anthropic's usual privacy stance, made in the name of catching novel attacks.
Pricing and How to Access It
Fable 5 is powerful but not cheap. Both Fable 5 and Mythos 5 are priced at:
- $10 per million input tokens
- $50 per million output tokens
That is roughly double the price of Claude Opus 4.8, but less than half the cost of the earlier Mythos Preview. For most users, the higher price is justified only when the task genuinely benefits from Fable's extra reasoning and autonomy — routine work is still better served by cheaper models.
| Access Path | Details |
|---|---|
| Claude API | Available immediately, pay-per-token |
| Claude apps (Pro, Max, Team, Enterprise) | Free through June 22, 2026; usage credits required after |
| Cloud platforms | Available on Amazon Web Services with built-in safeguards |
| Developer tools | Cursor and GitHub added support at launch |
| Industry platforms | Legal AI platform Harvey integrated Fable 5 |
Why Fable 5 Matters
Step back from the specs and three bigger themes stand out.
- The capability ceiling just rose for everyone. Tasks that were out of reach — like a company-wide code migration in a day — are now things a single developer can trigger. Cursor's CEO said Fable 5 "opened up a class of long-horizon problems that were out of reach."
- Safety is becoming a product feature, not an afterthought. The fallback-to-Opus design is a genuinely new pattern: a model that quietly hands dangerous questions to a safer sibling. Expect rivals to copy it.
- The race-versus-caution tension is now in the open. Releasing your most powerful model days after warning the world about powerful models is the clearest sign yet of the pressure AI labs are under to ship while staying safe.
For everyday users and businesses, the practical takeaway is that the frontier of what AI can do for you — especially in coding, research, and analysis — just moved forward significantly, with guardrails that try to keep the downside contained.
Frequently Asked Questions
What is Claude Fable 5?
Claude Fable 5 is Anthropic's most powerful publicly available AI model, released on June 9, 2026. It is the first "Mythos-class" model offered to the general public — a capability tier above Claude Opus. Fable 5 is state-of-the-art across most benchmarks in software engineering, knowledge work, vision, and scientific research, and it ships with built-in safeguards that route sensitive queries to the safer Claude Opus 4.8 model.
What is the difference between Claude Fable 5 and Mythos 5?
Fable 5 and Mythos 5 are the same underlying model. Fable 5 is the public version with safety guardrails active, so it declines or reroutes high-risk requests in cybersecurity, biology, and chemistry. Mythos 5 is a restricted version with some safeguards lifted, available only to a small group of vetted cyberdefenders and infrastructure providers through Anthropic's trusted-access program.
Why did Anthropic release Fable 5 right after warning AI is too dangerous?
The underlying Mythos model was considered too dangerous for general release because of its ability to find and exploit software vulnerabilities, so it was limited to vetted partners. Fable 5 is Anthropic's attempt to give the public access to that power safely, by wrapping the model in conservative safety classifiers that block high-risk topics and fall back to a less capable model. The release just days after Anthropic urged the industry to slow down highlighted the tension between racing ahead and managing risk.
How much does Claude Fable 5 cost?
Claude Fable 5 (and Mythos 5) is priced at $10 per million input tokens and $50 per million output tokens — roughly double the price of Claude Opus 4.8, but less than half the cost of the earlier Mythos Preview. It was included free on Pro, Max, Team, and Enterprise subscription plans through June 22, 2026, after which it requires usage credits on those plans.
How powerful is Fable 5 compared to other AI models?
Fable 5 is state-of-the-art on nearly all tested benchmarks of AI capability. Stripe used it to complete a 50-million-line codebase migration in a single day that would have taken a team two months. It is the new leader in vision tasks, accelerated drug-design work about tenfold for Anthropic's internal scientists, and can work autonomously for longer than any previous Claude model.
What safety guardrails does Fable 5 have?
Fable 5 uses three main safety classifiers covering cybersecurity, biology and chemistry, and model distillation (anti-cloning). When a query trips a classifier, the response is handled by the safer Claude Opus 4.8 instead. These safeguards are intentionally conservative and trigger in fewer than 5% of sessions. Anthropic also ran a 1,000-hour external bug bounty that found no universal jailbreaks, and it enforces 30-day data retention on all Mythos-class traffic for safety monitoring.
Can the public use Mythos 5?
No. Mythos 5 is restricted to a small set of vetted cybersecurity partners under Anthropic's "Project Glasswing" program, with plans to extend access to selected biomedical researchers. The general public can only use the safeguarded Fable 5 version.
Where can I use Claude Fable 5?
Fable 5 is available through the Claude API, the Claude apps, consumption-based Enterprise plans, and cloud platforms including Amazon Web Services. Many developer tools such as Cursor and GitHub added support at launch, and legal AI platform Harvey integrated it as well.
Final Thoughts
Claude Fable 5 is the clearest snapshot yet of where AI stands in 2026: astonishingly capable, commercially irresistible, and genuinely unnerving enough that the company shipping it spent the prior week warning everyone to be careful. Anthropic's bet is that the right answer is not to hide this power but to release it behind smart guardrails — letting a developer migrate a million lines of code in a day while a safety net quietly catches the questions that should not be answered.
Whether that bet holds up will depend on how well those guardrails survive contact with millions of real users. For now, Fable 5 raises the ceiling on what AI can do for ordinary work — and turns the industry's central tension, speed versus safety, into something you can watch play out in real time.
We will keep covering the AI frontier as it moves. Bookmark SaveDelete for clear, no-hype breakdowns of every major model launch.