claude-fable-5 — pricing, context window and limits

The most capable model in the current lineup, priced accordingly

Fable 5 sits at the top of the current lineup on capability, and its price reflects that positioning — it's the most expensive current model per token, on both input and output. Thinking runs on every request here with no way to turn it off; sending an explicit thinking configuration is rejected outright rather than adjusted or ignored, which is a genuine behavioural difference from every other model in the lineup, not a minor implementation detail.

The retention requirement that rules it out for some organisations

Fable 5 requires a short data-retention window to operate, which makes it structurally unavailable to any organisation running under a zero-data-retention policy — not a configurable option, a hard incompatibility. See Fable 5 is unavailable to zero-data-retention organisations for what that looks like in practice and what to route to instead.

Platform availability is the open question here

We could not confirm this model's availability on the major cloud-hosted platforms beyond the first-party API, so no claim is made either way — check current platform documentation directly before assuming coverage that isn't confirmed here.

Related

See claude-mythos-5 for the functionally identical model reachable through a separate, gated access path, and Opus 5 vs Fable 5 for how this model's price and capability compare directly against the mid-tier alternative most teams end up weighing it against.