claude-fable-5 is still active at Anthropic; Sume runs it as Fable 5.1

On Anthropic's side, claude-fable-5 is still an active model, with a retirement date not sooner than June 9, 2027. On Sume's side, the id is a retired spelling: a Format run that sends claude-fable-5 as model runs on Claude Fable 5.1, and the receipt echoes the id that ran. The two systems disagree about whether the model is retired, and for a Format run the Sume catalog decides.

That difference is easy to miss, because a direct call to Anthropic with the old id keeps working while the same id in a Format quietly lands on the newer model.

What does the vendor page say?

The models overview lists Fable 5.1 as claude-fable-5-1 at $10 input and $50 output per 1M tokens, with a 1M context window and 128K maximum output. The deprecations page lists the earlier model separately.

How does Sume treat the old id?

In the catalog code, the Fable 5 row names Fable 5.1 as its successor, and the successor is marked so that it resolves even when the workspace's extended catalog is closed. The result is that claude-fable-5 is accepted and runs on 5.1, while the plain claude-fable-5-1 spelling is accepted only where the catalog lists the row, because Fable 5.1 is otherwise a gated row. I checked these by calling the resolver in the repository, not by running a paid Format.

So the old id can work where the new one does not. If you want a predictable result, send the id your workspace's catalog lists and read the echoed model.

Is Fable worth it for a Format?

Price is the plain trade-off: Fable 5.1 lists at 2.5 times Opus 5.5's input rate and output rate on the vendor page ($10 against $4 input, $50 against $20 output). Whether the extra capability pays for a particular Format is a test you can run on your own input, comparing the two receipts. The runs doc explains usage.debited_usd_micros, which includes the turn's own LLM row.

What should I do with a saved id?

Each choice trades predictability for tracking, and the receipt is how you see which one you got.

• If a Format saves claude-fable-5, expect Fable 5.1 to run it and read the receipt to confirm.
• If you need the older behavior exactly, call Anthropic directly while its date allows it.
• If you want to avoid the question, omit model and let the default apply.

Why does the difference matter for cost?

If a team saved the old id in a Format and compared receipts over time, a move to Fable 5.1 would appear as a change in the model echoed on the receipt rather than as a change in the Format. That is the place to look when a run costs more or behaves differently than last month's. The id in your saved call stays the same; the model that ran does not.

Anthropic's models overview lists the 5.1 model's thinking as adaptive and always on, so there is no switch to turn reasoning off for it in a direct call. The Sume picker documents its own effort levels separately, so read the picker and the receipt for what a Format run used, not the vendor's defaults.

How do I confirm which model ran?

Read the receipt. The call doc says it echoes the id that ran, so a saved claude-fable-5 should show the 5.1 id back. If it does not, the id was handled differently on your workspace and you should not assume the mapping above. Test with one run, not a bulk list.

Keep the vendor's date in mind only for direct calls. On Anthropic's API, claude-fable-5 stays addressable until its own retirement date, so a team that calls both can see the same id behave differently in two places.