What Is Claude Fable 5?
Claude Fable 5 is Anthropic’s most capable publicly available AI model, released on June 9, 2026. It is a “Mythos-class” model — the same underlying model as Claude Mythos 5 — made safe for general use through built-in safety classifiers that route high-risk queries to Claude Opus 4.8 instead.
That one-paragraph answer is the headline. The longer story is more interesting: Fable 5 marks the first time Anthropic has released a model from its Mythos tier — the class of models that sits above Opus in capability — to the general public. Until now, Mythos-class capabilities were restricted to a small group of cyberdefenders and infrastructure providers through Project Glasswing, a collaboration with the US government.
Anthropic says Fable 5’s capabilities “exceed those of any model we’ve ever made generally available,” with state-of-the-art results on nearly all tested benchmarks. The longer and more complex the task, the larger its lead over previous Claude models.
Claude Fable 5 at a Glance
| Feature | Details |
|---|---|
| Developer | Anthropic |
| Release date | June 9, 2026 |
| Model class | Mythos-class (above Opus) |
| API model string | claude-fable-5 |
| Pricing | $10 / million input tokens, $50 / million output tokens |
| Fallback model | Claude Opus 4.8 (for flagged queries) |
| Fallback frequency | Under 5% of sessions |
| Availability | Claude API, Amazon Bedrock, Vertex AI, Microsoft Foundry |
| Sibling model | Claude Mythos 5 (safeguards lifted, restricted access) |
What Can Claude Fable 5 Do?
Software engineering
This is where the gap with older models is widest. During early testing, Stripe reported that Fable 5 performed a codebase-wide migration in a 50-million-line Ruby codebase in one day — work that would have taken an entire engineering team over two months by hand. On Cognition’s FrontierCode evaluation, which measures whether models can complete hard coding tasks to production-codebase standards, Fable 5 scores highest among frontier models even at medium effort. Cursor called it the state of the art on CursorBench, and GitHub highlighted its reliability on long-horizon, autonomous coding tasks.
Knowledge work and analysis
Fable 5 posted the highest score of any model on Hebbia’s Finance Benchmark for senior-level reasoning, with gains in document-based reasoning, chart interpretation, and problem solving. Hex reported it was the first model to break 90% on its core analytics benchmark — a 10-point jump over Opus. Trading firm IMC said it “aced” their trading-analysis evaluations nearly across the board.
Vision
Fable 5 is the new state of the art for vision tasks. It can extract precise numbers from dense scientific figures and rebuild a web app’s source code from screenshots alone. In a striking demo, it completed Pokémon FireRed using only raw game screenshots — no maps, no navigation aids, no helper harness — something earlier Claude models couldn’t do even with extensive scaffolding.
Memory and long-running autonomy
The model stays focused across millions of tokens and improves its own outputs using persistent notes. In Anthropic’s Slay the Spire test, file-based memory improved Fable 5’s performance three times more than it did for Opus 4.8.
Science
The Mythos 5 variant accelerated parts of Anthropic’s internal drug-design process by roughly ten times, autonomously handling protein binding-site selection, tool runs, and failure recovery. In blinded comparisons, Anthropic scientists preferred its molecular biology hypotheses about 80% of the time over Opus-class models — and one of its hypotheses was independently corroborated by an outside lab.
How Do Claude Fable 5’s Safety Guardrails Work?
This is the defining feature of the release — and the reason “Fable” and “Mythos” are two different products built on one model.
Fable 5 ships with safety classifiers: separate AI systems that screen requests in three areas — cybersecurity, biology and chemistry, and model distillation. When a classifier triggers, the response is handled by Claude Opus 4.8 instead, and the user is told this happened. Anthropic’s early data shows more than 95% of sessions involve no fallback at all.
The safeguards are deliberately tuned to be conservative, so some benign requests will trip them — Anthropic acknowledges this and says reducing false positives is a post-launch priority. The robustness testing was substantial: an external bug bounty found no universal jailbreaks in over 1,000 hours of testing, and external red-teaming organizations also came up empty on long-form agentic tasks (though the UK AISI made partial progress in a brief window).
One consequential policy change accompanies the launch: all Mythos-class traffic requires 30-day data retention, even for enterprises that previously had zero-retention agreements. Anthropic says the data won’t be used for training — only for detecting novel attacks and reducing false positives.
Claude Fable 5 vs Claude Mythos 5: What’s the Difference?
They are the same underlying model. The difference is the safeguards. Fable 5 has all classifiers active and is available to everyone. Mythos 5 has cyber safeguards lifted and is restricted to vetted partners — initially Project Glasswing cyberdefense organizations, with a trusted access program for biology researchers coming next. The naming is deliberate: fable (from the Latin fabula) and mythos (Greek) both mean “that which is told.”
Claude Fable 5 Pricing and Availability
Both Fable 5 and Mythos 5 cost $10 per million input tokens and $50 per million output tokens — double Opus 4.8’s price, but less than half the price of the earlier Claude Mythos Preview.
Access rolls out in stages:
- Now: Fully available on the Claude API and consumption-based Enterprise plans, plus Amazon Bedrock, Vertex AI, and Microsoft Foundry.
- Through June 22, 2026: Included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
- From June 23, 2026: Removed from subscription plans; usage credits required. Anthropic says it aims to restore subscription access once capacity allows.
Why Claude Fable 5 Matters
The release answers a question the AI industry has been circling for a year: how do you ship a frontier model whose raw capabilities are too dangerous to release unfiltered? Anthropic’s answer — one model, two products, classifier-enforced fallback — is the first large-scale test of that approach. The timing is notable too: it arrived just days after Anthropic publicly urged major AI labs to establish a coordinated “brake pedal” on frontier development amid concerns about recursive self-improvement, and as the company prepares for public markets alongside OpenAI.
Whether the safeguard model becomes the industry template — and whether users tolerate occasional false positives and mandatory data retention as the price of frontier access — will play out over the coming months.
New to the underlying tech? See our explainer on what a Large Language Model (LLM) is, and how models like this power the best AI coding tools in 2026.
Frequently Asked Questions
What is Claude Fable 5?
Claude Fable 5 is Anthropic’s most powerful publicly available AI model, released June 9, 2026. It’s a Mythos-class model with built-in safety classifiers covering cybersecurity, biology/chemistry, and distillation.
How much does Claude Fable 5 cost?
$10 per million input tokens and $50 per million output tokens via the Claude API — about twice the price of Claude Opus 4.8.
Is Claude Fable 5 better than Claude Opus 4.8?
Yes, on nearly all tested benchmarks — particularly long, complex tasks in coding, analysis, and vision. Opus 4.8 remains the fallback model when Fable 5’s safety classifiers trigger.
What’s the difference between Claude Fable 5 and Claude Mythos 5?
Same model, different safeguards. Fable 5 is publicly available with full safeguards; Mythos 5 has cyber safeguards lifted and is restricted to vetted partners.
How often do Fable 5’s safeguards trigger?
In less than 5% of sessions, per Anthropic’s early data. Flagged queries get a response from Claude Opus 4.8 rather than an outright refusal.
Can I use Claude Fable 5 on a Claude Pro or Max plan?
Yes, at no extra cost through June 22, 2026. After June 23, it requires usage credits until Anthropic restores standard subscription access.
