The Claude Fable family
This family has had two members since 1 September 2026, Fable 5 and Fable 5.1, and both charge $10 in and $50 out, exactly double Opus 5, while both carry a one million token context window and both cap output at 128K tokens. What the extra money buys is not in the spec table.
- double the Opus 5 price
- adaptive thinking always on
- the dearest Claude tier
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. What changed
So where is the difference
Nowhere in the spec table, and that is the point of this page. Put Fable 5 and Opus 5 side by side on window, maximum output, delivery routes and even knowledge cutoff: a one million token window on both, 128K output on both, and both shipping on the same five routes, from Anthropic own API through Bedrock, Claude Platform on AWS, Google Cloud and Microsoft Foundry. Fable 5 has a January 2026 knowledge cutoff against May 2026 for Opus 5, so on that one column the dearer model is the older one.
The difference lives in two other columns. Anthropic calls Fable 5 its most capable widely released model and describes it as next-generation intelligence for long-running agents. And its adaptive thinking column reads yes, always on, where Opus 5 and Sonnet 5 read simply yes.
Two leaderboard numbers that point opposite ways
This is where the real recommendation comes from. On the Arena WebDev leaderboard, Fable 5 scores 1628 against 1692 for Opus 5. On the general text leaderboard it reverses: Fable 5 sits at 1506 against 1494 for Opus 5.
The gap is sixty-four points on code and twelve points on text. So most capable does not mean best at everything, and if your work is code then the model charging double sits lower at exactly that job. Our coding table shows the same thing: Fable 5 second, Opus 5 first, 15.4 points apart.
Two things the extra money loses you
First, speed. In the Anthropic table the comparative latency column reads slower for Fable 5, moderate for Opus 5 and fast for Sonnet 5. It is the only model in that table marked slower. For an agent grinding overnight that does not matter. For anything a person waits on, it does.
Second, long batched output. Anthropic runs a beta on the Message Batches API that lifts the output ceiling to 300K tokens, and it names the supported models explicitly: Opus 5, Opus 4.8, 4.7, 4.6, Sonnet 5 and Sonnet 4.6. Fable 5 is not on that list. So if your job is producing very long output in batches, the cheaper model does something the dearer one does not.
Thinking you cannot switch off
Adaptive thinking being always on carries a financial consequence that the unit price hides. Thinking tokens are output tokens, and output here costs $50. On Opus 5 you can lower the effort level; here you do not have that lever.
One more thing moves the invoice, and it belongs to the whole generation rather than to Fable alone: the tokenizer introduced with Opus 4.7 turns the same text into roughly thirty percent more tokens. Fable 5 carries it too. So comparing the paper price of two models from different generations understates the bill you actually get.
The second member, and the two you cannot buy
Claude Fable 5.1 arrived on 1 September 2026 and made this a two-member family. Its base price matches Fable 5 and so do the context window and the output cap. Three things changed: cache reads went from one dollar to twenty-five cents, the knowledge cutoff moved from January to June 2026, and benchmark scores rose on long agentic work. The full arithmetic and its boundaries are on the Fable 5.1 page.
Two further members cannot be bought at all. Claude Mythos 5 shares Fable 5 specs and pricing and shipped on the same day, 9 June 2026, and Mythos 5.1 repeated that on the day Fable 5.1 launched. Neither is generally available: both are offered by invitation inside Project Glasswing, with no self-serve sign-up. We do not count them as buyable members of this family, because something you cannot buy has no place on a page about what to buy.
Specifications
Every number here comes from the vendor's own page, with that page linked beside it.
Release timeline
Each version with its own release date, and what changed against the one before it.
- Claude Fable 5.1 current Base price unchanged, cache at a quarter, and the knowledge cutoff now sits ahead of Opus 5.
- Claude Fable 5 legacy Anthropic dearest widely released model: first at general conversation, second at code.
Using it from Iran
This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.
- Reachable
- blocked
- Payment
- no working route
- Free tier
- no
- Measured on
How we checked: Iran is not on Anthropic supported-countries list. Read from Anthropic own page rather than measured on the network.
What it is good at
- The highest Claude score on the Arena general text leaderboard, 1506 against 1494 for Opus 5
- The same one million token window and the same 128K output as the Claude upper tier
- Available on all five delivery routes, from Anthropic own API to Microsoft Foundry
- Adaptive thinking is always on, so its behaviour is more uniform across requests
Where it falls short
- It costs twice Opus 5 while the context window, maximum output and delivery routes are identical.
- It sits below Opus 5 on the WebDev leaderboard, 1628 against 1692, so for code the dearer model is the weaker one.
- It is the only model in the Anthropic table whose comparative latency column reads slower.
- The 300K output beta on the Message Batches API does not cover Fable 5, while it covers Opus 5 and four other models.
- Adaptive thinking is always on, so you do not have the effort lever for bringing the cost down.
- Its knowledge cutoff is January 2026, four months behind the May 2026 cutoff of Opus 5.
- Iran is not on the Anthropic supported-countries list, so there is no official route.
Our take
If your work is code, take Opus 5: it scores higher and costs half. Justify Fable 5 only when your job is the one they built it for, long-running agents and general text, and when neither speed nor long batched output is your problem.
Questions people actually ask
Is Fable 5 better than Opus 5
It depends on the job, and the two Arena numbers point opposite ways. Opus 5 leads on WebDev at 1692 while Fable 5 leads on general text at 1506. Given that Fable costs double, the clear choice for code is Opus 5.
How many versions are in the Fable family
One, Fable 5, generally available since 9 June 2026. Claude Mythos 5 carries the same specs and the same price but is offered by invitation inside Project Glasswing with no self-serve sign-up, so it is not a buyable member of this family.