Claude Fable 5 Is Free on Paid Plans Until June 22. Here Is What to Do With the Window.

On June 9, Anthropic released Claude Fable 5, which it describes as its most capable model ever made generally available (announcement). The headlines mostly argued about what it means for the AI race. If you run a business, the practical part is simpler and has a date attached.

Fable 5 is included at no extra cost on paid Claude plans, Pro, Max, Team, and seat-based Enterprise, through June 22, 2026. Starting June 23, it comes out of those plans and using it requires usage credits, until Anthropic says capacity allows it to be included as standard again (Anthropic's announcement).

That gives you roughly two weeks to find out, for free, whether the strongest model on the market changes anything for your work. Here is how to think about it.

First, the one distinction that most coverage missed

The June 23 change applies to subscription plans, the Claude apps you log into with a monthly plan. It does not apply to API billing, where software calls Claude directly and pays per use. API access to Fable 5 is available normally and is not part of the free-window mechanics (Anthropic docs).

So: if you or your team use Claude through a Pro, Max, Team, or Enterprise plan, the window applies to you. If your tools call Claude through the API, nothing about your billing changes on June 23. The per-use price for Fable 5 on the API is $10 per million input tokens and $50 per million output tokens, which is double the price of Claude Opus 4.8 (pricing docs). More on what that buys in a moment.

One honest gap: Anthropic has not published exactly what Fable 5 usage will cost subscription users after June 23, beyond saying it will require usage credits. Treat any specific number you see elsewhere as a guess.

What is actually new about Fable 5

In Anthropic's words, Fable 5 is a "Mythos-class model made safe for general use," and its capabilities exceed anything the company has released broadly before. It reports state-of-the-art results across coding, knowledge work, and research benchmarks, and early customers describe large multi-month projects finishing in days (announcement). Those are vendor and customer claims, not independent audits, so the sensible posture is curiosity rather than belief. The free window exists so you can check against your own work, which is the only benchmark that matters for you.

One design detail worth knowing: Fable 5 ships with built-in safety filters. For a small share of requests in sensitive areas, it declines and a different Claude model answers instead. Anthropic says more than 95 percent of sessions never hit this. You mostly will not notice, but if a specific task seems to get a different quality of answer, that can be why.

A simple plan for the next two weeks

Do not "try the new AI" by chatting with it aimlessly. You will learn nothing transferable. Instead:

1. Pick your three most annoying recurring tasks. The proposal that takes a full afternoon. The spreadsheet cleanup nobody wants. The report you assemble from four sources every month.
2. Run each one through Fable 5 once, for real. Use an actual example with real stakes, not a toy version. In the Claude apps, select Fable 5 as your model while the window is open.
3. Compare against your current way honestly. Time saved, quality, how much fixing the output needed. Write the result down, even one line.

By June 22 you will know, from evidence instead of headlines, whether the frontier model earns a place in your operation or whether the standard models already cover your needs. Both answers are wins. Knowing is the win.

That sorting exercise, deciding which work deserves the strongest tool, is the same map we build with operators every week. If the window test surfaces something promising and you want a second set of eyes on it, you know where to find us.