Claude Fable 5: The Top-Tier Model for Business

New model · June 9, 2026

Anthropic's most powerful model, now open to everyone. An honest breakdown: how it differs from the secret Mythos and from Opus 4.8, where it actually pays for itself, and where you shouldn't overpay for it. Inside: 4 scenarios with ready-to-use prompts.

⏳ The window closes June 22

This model spent a year under lock and key: access was limited to corporations through a closed program. Now, for the first time, it's open to everyone.

Until June 22, Fable 5 is included in the Pro, Max, and Team subscriptions. From June 23, it will only be available via the API, at steep pay-per-use rates. That's two weeks to test-drive the top model firsthand and figure out where it can strengthen your business before it gets expensive.

01 · What this actually is

Mythos, Fable, and how they differ.

A year ago, Anthropic built Mythos, a model so powerful they locked it away immediately. Mythos autonomously finds security holes in any system: a 73% success rate on expert-level hacking tasks, thousands of vulnerabilities in Windows, macOS, and browsers. Access went only to vetted partners through a secret program.

Mythos 5: the closed one

Full power, no restrictions. Corporations and special-access partners only. The very "technology that scared governments."

Fable 5: the public one

The same engine as Mythos, plus safety catches. Dangerous topics (hacking, biology) get automatically routed to the more cautious Opus 4.8. Everything else runs at full power.

In plain terms

Fable is Mythos with a muzzle on the dangerous stuff so it could be released to the public. For business tasks, you'll never notice it.

02 · Performance

Where Fable beats everything else.

It's not "slightly better across the board." There are whole classes of tasks where Opus 4.8 fails outright and Fable pushes through. Especially code, vision, and long autonomous work.

The paradox

Fable is slower than Opus (minutes versus seconds), and that's exactly why it's smarter. The extra time goes into deeper search.

03 · What it costs

The price per million "chunks of text."

A token is a chunk of text used for billing. One page ≈ 1,500–2,000 tokens. Fable is the most expensive: twice the price of Opus, and 4–5x the price of the cheap models.

The number one money-saving rule

On simple text (an email, a post, a short reply), Fable is no better than the cheap models, just pricier. It's only worth paying for when there's an image involved, a large volume of data, or a high cost of error.

04 · Working in tandem

The expensive one builds; the cheap one runs it.

The single best budget-saving move: don't put Fable on everything. Run two models as a team.

Fable for the heavy lifting

Vision, large volumes, complex architecture, building from scratch. One expensive pass, done right.

The cheap one for routine

Daily replies, simple copy, operating a tool that's already built. Pennies per request.

The routing rule

Call in the expensive model only for the heavy stuff: vision, large volumes, building from scratch. Hand routine work to the cheap one. This one habit keeps costs under control and stops you from overpaying for nothing.

05 · One run, maximum value

4 tasks where a single session leaves you with a lasting result.

Imagine you get one precious run of the expensive model. Spend it on something that leaves a finished artifact in your hands: a design, code, a tool, a plan. The scenarios go from the flashiest to the most foundational. Copy the prompt and fill in your details in [brackets].

01 · Drop in a great website → walk away with the design

Superpower: Vision

Find a site whose design you love: a competitor, a big-name service, any reference. Drop in a screenshot. Fable doesn't just look at the picture; it parses the layout's structure and builds working code for the same page, adapted to your product.

Prompt: copy and paste into Claude:

Here's a screenshot of a website whose design I like. Build working HTML and CSS for the same page: responsive on phone and desktop, clean code. Replace the copy and products with mine: [paste yours]. Set it up so I can change the colors and content myself, and explain where everything lives.

On a cheap model: A cheap model will build something that "looks close-ish": the layout will drift, blocks will get shuffled, and on a complex design it turns to mush.

Why it's worth it: Agency-grade design built in an evening. Yours forever, unlike one-off advice.

02 · Describe an idea → get a working product

Superpower: Development + stamina

You describe what you want in plain words. Fable writes code for hours, holding the entire architecture in its head, and builds the first working version: a bot, a mini-service, an internal tool, a landing page with logic.

Prompt: copy and paste into Claude:

Build a working first version of this: [describe your product in your own words]. I need complete code that actually runs, not fragments. Keep the entire architecture in view and don't lose the connections between parts. Finish with step-by-step instructions for running it on my end.

On a cheap model: A cheap model will build the first few screens, then lose the thread, duplicate code, and break what it already wrote. On a big project it falls apart.

Why it's worth it: One session = a prototype you won't be embarrassed to show an investor or a client. It stays with you as the foundation of your product.

03 · Take someone else's project → make it your tool

Superpower: Stamina on unfamiliar code

GitHub holds thousands of free projects that solve 80% of your problem. Pick a good fit, and Fable holds the entire unfamiliar codebase in its head and reworks it for you.

Prompt: copy and paste into Claude:

Here's an existing project: [GitHub link or paste the code]. Adapt it to my task: [what exactly needs changing and adding]. Change the logic where needed without breaking the rest. First give me a brief explanation of how the project is structured, then show what you're changing and where.

On a cheap model: A cheap model gets lost in a large unfamiliar codebase: it changes one thing, breaks three others, and misses the dependencies.

Why it's worth it: Months of someone else's work become your tool in one evening. And they stay yours.

04 · A big project idea → a map for building it

Superpower: Depth

You have an idea for a big service (a marketplace, a niche CRM, an aggregator). Describe it in a couple of paragraphs and you get a complete map: what it's made of, how it all connects, what order to build in, and where it will blow up.

Prompt: copy and paste into Claude:

I want to build this service: [describe your idea in your own words, 2–3 paragraphs]. Think through the full architecture: what parts it consists of, how they connect, what data lives where, what order to build everything in, and where the future bottlenecks and risks are. Give me a document: modules, data schema, development stages, and the technical forks in the road with pros and cons.

On a cheap model: A cheap model gives you a generic plan off the internet: it misses your project's specifics, doesn't catch the pitfalls, and contradicts itself over a long plan.

Why it's worth it: An architect's work that would run you 100–200K. A mistake in the foundation = rewriting everything six months later. The document is yours forever.

✕ Don't waste your run

Simple stuff that Cowork or a cheap model can handle.

✓ Save it for this

Hard stuff that needs the full power all at once.

06 · Before you start

Your data, and a first step tonight.

Data safety

Anthropic doesn't train on your data and keeps logs for just 7 days. Still, a little caution won't hurt: anonymize real customer personal data, ID documents, and financials before uploading. Contracts and reports with no personal data in them are usually safe to upload.

Try it in 15 minutes

• 01 — Open Claude and select the Fable 5 model.
• 02 — Find a website whose design you like and take a screenshot.
• 03 — Paste the prompt from scenario 1 and attach the screenshot.
• 04 — A couple of minutes later, you're holding working code for that design, adapted to your product.

If your reaction is "this would have saved me half a day," you need this model. If not, you've spent 15 minutes and found out before paying for the expensive API.

What's next

The course: an AI team for your business.

Fable 5 is one strong model. In the course, we show you how to build an entire AI team: agents that run your social media, answer customers, and build websites, without your daily involvement. Put together your own AI team that works for you every day.