Claude Fable 5 Pricing: What It Costs and How to Estimate

Anthropic shipped Claude Fable 5 on June 9, 2026, and it's the first Mythos-class model anyone outside the lab can actually use. The headline question for most people isn't what it can do, it's what it costs. So let's get straight to the numbers, then walk through how to estimate your own bill without a spreadsheet or a math degree.

Quick note before we start: we run ShadowProof, an independent app. We're not affiliated with Anthropic, and nothing here is official guidance from them. These are the published facts plus some plain-English math to help you plan.

The per-token price

Claude Fable 5 costs '$10 per million input tokens' and '$50 per million output tokens' on the API. The model id is 'claude-fable-5'.

Input tokens are everything you send in: your prompt, your system instructions, any documents or context you paste. Output tokens are what the model writes back. Output is five times the price of input, which is normal for these models. The model does more work generating text than reading it, and that gap shows up in the bill.

Two prices, one model. That's the whole rate card for raw API usage. Whether that feels cheap or steep depends entirely on what you're comparing it to, so let's do that next.

How Fable 5 compares

Here's the part that surprises people. Fable 5 is exactly double the price of Claude Opus 4.8. If Opus 4.8 was your baseline, you're looking at twice the per-token cost for the same volume of work.

But it cuts the other way too. Fable 5 costs less than half what the earlier Claude Mythos Preview ran. So if you'd been paying for that preview, this is a real drop, not a hike.

Why it costs more than Opus

The price isn't arbitrary. Anthropic positions Fable 5 as more capable than Opus 4.8, and they point to a few areas in particular:

• Long-horizon, complex tasks that span many steps
• Coding
• Finance work
• Vision

If your work lives in those buckets, the higher rate can pay for itself by getting more done per call. If you're doing simpler, short tasks, you might not need Fable 5 at all, and Opus 4.8 at half the price could be the smarter pick. Match the model to the job.

The free window (read the fine print)

Here's the good news if you don't want to touch the API yet. From June 9 to June 22, 2026, Claude Fable 5 is included at no extra cost in the Pro, Max, Team, and seat-based Enterprise subscription plans. If you already pay for one of those, you can use the model in that window without paying per token.

After June 23, 2026, access through those plans needs usage credits instead. Anthropic has said it plans to restore broader access once capacity allows, but there's no firm date attached to that promise.

I want to be honest about this: that free window is short and it's time-limited. Plans like this can shift, get extended, or get pulled early depending on demand and capacity. Treat the June 22 date as the plan today, not a guarantee. If free access matters to you, use it while it's there and don't build a workflow that assumes it'll last.

For the always-on routes, Fable 5 is fully available right away on the Claude API, on consumption-based Enterprise plans, and on AWS Bedrock. Those bill by usage, free window or not.

A plain example of what a task costs

Token counts feel abstract until you turn them into dollars, so let's ground it.

A rough rule of thumb: '1 million tokens' is about 750,000 words. That's an estimate, not an exact conversion, since tokenization varies by language and content, but it's close enough to plan with.

Work it backward. Say you ask Fable 5 to write a 1,500-word article. That output is roughly 2,000 tokens. At '$50 per million output tokens', 2,000 tokens costs about '$0.10'. Ten cents for a 1,500-word draft.

What that estimate leaves out

That dime only covers the output. Your real cost is input plus output. So if you fed the model a long brief, a few reference documents, and a detailed system prompt, you'd add input tokens at '$10 per million' on top of that ten cents. For a short prompt the input cost is tiny. For a request that includes a 50-page PDF, the input side can easily become the bigger number.

So the honest takeaway: a single, lightly-prompted 1,500-word output runs around '$0.10', call it a rough estimate, and your actual bill scales with both how much you send and how much you get back.

Scaling it up

If ten cents is one article, then a hundred similar outputs is somewhere near '$10' on the output side, again before input. That's the kind of back-of-the-envelope math worth doing before you wire Fable 5 into anything that runs at volume. Multiply your typical output length by your call count, convert to tokens, and you'll have a usable estimate in about a minute.

Want more on the model itself or the free question? We dig into both in our Claude Fable 5 explainer and in is Claude Fable 5 free.

Should you pay for it?

It depends on the work, and I mean that literally rather than as a dodge. If you're doing long, multi-step projects, serious coding, finance analysis, or anything visual, the extra capability over Opus 4.8 can justify paying double. If your tasks are short and simple, the cheaper model probably gets you there for less.

If you've got a Pro, Max, Team, or seat-based Enterprise plan right now, the smartest move is to try Fable 5 during the free window and see whether it actually changes your output. Real testing beats a pricing table every time. Then decide whether the per-token rate is worth it once the window closes.

Frequently asked questions

How much does Claude Fable 5 cost?

On the API, Claude Fable 5 costs '$10 per million input tokens' and '$50 per million output tokens'. Those are the raw usage rates. The model is also included in some subscription plans during a limited free window from June 9 to 22, 2026.

Is Claude Fable 5 more expensive than Opus 4.8?

Yes. Fable 5 is double the price of Claude Opus 4.8 per token. It's also less than half the price of the earlier Claude Mythos Preview. Anthropic prices it higher than Opus because it's more capable on long, complex tasks, coding, finance, and vision. We compare them more in Claude Fable 5 vs Opus 4.8.

Is there a free way to use Claude Fable 5?

For a limited time, yes. From June 9 to 22, 2026, it's included at no extra cost in the Pro, Max, Team, and seat-based Enterprise plans. After June 23 you'll need usage credits. That window is short and could change, so don't count on it lasting.

How do I estimate my cost?

Estimate the words you expect, convert to tokens (about '1 million tokens' per 750,000 words), then apply the rates. A 1,500-word output is roughly 2,000 tokens, or about '$0.10' at the output rate. Add your input tokens at '$10 per million' on top. Treat the result as an estimate, since real cost depends on input plus output.