Claude Fable 5.1

What Claude Fable 5.1 is

Claude Fable 5.1 is a commercial cloud model for demanding reasoning and long-horizon work. Long-horizon describes work with multiple dependent steps, where earlier decisions affect later actions. The model is accessed through supported cloud services rather than downloaded as local weights.

The provider generally suggests starting with Opus 5.5 for most workloads. Fable's place in the catalog is therefore a specific evaluation option, not a claim that a more expensive or slower model is automatically better. A difficult task needs a clear success criterion before the extra effort can be judged useful.

How it fits into a workflow

An illustrative evaluation gives the model a multi-part problem with known constraints and asks for a reviewable plan followed by a bounded result. Record whether it preserves those constraints across steps and how many corrections are required. If tools are involved, inspect their actions separately from the model's written reasoning.

Examples described as illustrative or proposed trials are editorial suggestions. Documented real builds, when available, are linked separately below.

Requirements, cost and limitations

Review current access, pricing, model-specific behavior and migration notes. The documentation lists text/image input, text output and large context/output limits. Those limits do not establish reliable recall of every detail or unlimited unattended work, and the application still owns tool access and runtime behavior.

When to choose it and what to compare

Consider Fable when a task-specific evaluation gives a reason to go beyond your usual model. Use the same inputs and acceptance checks across candidates, and include delay and total usage in the decision. Keep a simpler model or non-agent process when it already meets the requirement.

Access, running cost and practical requirements

Frontier hosted candidate — editorial selection. How we use this label.

Claude API and listed partner platforms; selectable in eligible Claude products. Exact subscription entitlement was not established in this bounded audit.

Developer API prices

Billing route

Claude API base

Units

USD per one million tokens

Input

$10

Cached input

$0.25

Output

$50

5-minute cache write

$12.5

1-hour cache write

$20

Batch input/output discounted 50%. Current API prices do not establish lower total task cost than Opus.

API prices are not a chat subscription price. A token is a unit of processed text; images and other inputs have their own token accounting. Repeated history, reasoning output and tools can change the final bill. A subscription may offer separately claimed credits, subject to its terms.

What the configuration changes

Provider recommends trying it when higher-effort Opus evaluations still fail the requirement.

Adaptive thinking always on; API default high.

Forced tool choice is rejected; thinking-state compatibility and edited conversation history need migration handling.

Higher unit cost and longer runs must be justified by a concrete successful outcome.

Benchmark evidence and what it means

These are selected source-reported observations, not a league table. Task, version, tool access, reasoning budget and test harness all affect a result. A historical score does not establish current leadership or accuracy on your own work.

Sources and evidence

Product facts are based on the following published sources, checked 2026-10-27. Workflow suggestions and comparisons are editorial interpretation; PrisAIm has not independently benchmarked this product.