Claude Fable 5: Pricing and Data Retention

Superseded by Claude Fable 5.1Anthropic's highest-priced Cursor model for long, complex work at $10 per million input tokens and $50 per million output, with $0.25 cache reads; Enterprise, Teams and Privacy Mode accounts need data-retention approval to use it. Press Enter for the full definition.. Cursor now recommends Claude Fable 5.1 in place of Fable 5. Input and output rates stay at $10 and $50 per million tokens, prompt-cache reads fall to $0.25 per million (75% below Fable 5), and the same data-retention approval applies. Cursor hides Fable 5 by default. For new work, read our Fable 5.1 page and the current model guide.

Cursor's Fable 5 page describes it as Anthropic's most capable generally available model for autonomous knowledge work and coding. It is a Mythos-class model built for long-running, complex and asynchronous work. Its stated advantage is holding intent across a long task with fewer check-ins than earlier models needed.

The same page puts Fable 5 at the top of CursorBenchCursor's internal benchmark that scores models on both task performance and token efficiency, not accuracy alone. Press Enter for the full definition. alongside Opus 5. That ranking is out of date: Cursor's Opus 5.5 page says that at Opus 5.5's launch, no model in Cursor had a higher CursorBench score.

The price and the retention terms make Fable an escalation model rather than a casual default. Cursor says Fable 5 costs about twice as much as Opus 5 and pitches it at a team's most complex work. Before you standardize on a Fable model, pick a task with a costly handoff or a long autonomous run, define the evidence you expect, and compare the finished result against Opus.

Fable 5 comes with a retention exception. Cursor states that Anthropic stores agent input and output data for automated and human harm-prevention processes, whatever your Cursor Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition. setting. Retained customer data is deleted after 30 days, automatically and permanently, unless a safety investigation or legal hold applies. By default no person can read it, and it is not used to train Anthropic models or products.

Cursor adds that retained customer data is not shared with other customers. Enabling Fable doesn't change the account's Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition. or Cursor's own retention. Privacy Mode's no-training commitment stays in place, and this model adds provider retention for the stated safety purpose.

Deletion after 30 days is still retention. If a repository or customer agreement allows only zero-retention providers, Fable doesn't fit that policy, even though training is ruled out. Keep the model disabled for that scope and choose an allowed model, instead of trying to keep sensitive strings out through prompting alone.

Customers with Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition. enabled must approve Fable 5's Data Retention Policy in the Cursor Dashboard before they can use the model. On Teams, that approval applies to the whole team. On Enterprise, availability is approved at the organization level, and team and group model access controls keep deciding who can select it.

Teams with Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition. enabled, and every Enterprise customer whatever its Privacy Mode setting, start with Fable 5 switched off. Individual customers and Teams with Privacy Mode disabled get it on by default.

That default only governs availability. It says nothing about whether the retention terms fit a company's policy.

1. 1Read the current Fable data-retention policy and map it to the data allowed in the target repositories.
2. 2Approve the policy at the required account, team, or organization level in the Cursor Dashboard.
3. 3For Enterprise, restrict model access to the intended teams or groups after organization approval.
4. 4Confirm the model appears in a fresh Cursor session and run a non-sensitive test task.
5. 5Record the approver, the policy version or review date, the scope, and the next review date.

Give team-wide acceptance a named owner, because one approval changes availability for everyone on the team. Enterprise group controls can narrow use afterward, but the organization has still accepted the provider terms. Review the decision whenever Anthropic or Cursor changes the policy, so an old dashboard state never stands in for current legal approval.

Cursor documents an automatic fallback for requests that trip Fable 5's security guardrails. Such a request goes to Claude Opus, so the work continues without a manual retry or model switch. A session you started on Fable can therefore contain work that Opus produced after the guardrail event.

The fallback keeps the work moving. It doesn't mean the two models share behavior, rates or approval status.

If model identity matters for an audit, a cost model or an evaluation, inspect the usage record and note which model ran. Cursor's model page describes the fallback, but it doesn't promise that a governed workflow can ignore the model change.

Cursor lists $10 per million input tokens, $12.50 cache write, $1 cache read and $50 output, twice the matching Opus 5 rates. Fable 5 is a third-party model, so The X SubscriptionThe subscription Cursor is sold through: one plan for Grok, Cursor and X, with Free, Plus, Premium, Super and Ultra tiers, per-seat Teams, a weekly included pool for Cursor models shared with Grok, and third-party models billed at API list price from prepaid credit. Press Enter for the full definition. charges it to prepaid credit, and legacy plans count it against the Other ModelsOn legacy Cursor plans, the monthly pool for supported third-party models from providers such as Anthropic, Google and OpenAI, separate from Cursor Models. On The X Subscription third-party models are billed from prepaid credit instead. Press Enter for the full definition. pool. The high thinking variant, which Cursor recommends for the strongest results, can also spend more tokens on a hard request.

Measure cost per verified task. A long Fable run can be economical if it completes work that would take several Opus retries and human interventions. It is expensive on a small change that ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition. or Sonnet can pass in one attempt. Weigh both the outcome and the supervision time on your own task set.

For new work, use Fable 5.1, which Cursor now recommends at the same input and output rates with cheaper cache reads. The rest of this section applies to either model. Use Fable for a long, asynchronous task with a clear finish line and data that the retention policy allows. Good candidates include a hard cross-system migration, a broad investigation that must keep a set of constraints intact, or a background build where repeated human check-ins are the main cost. Give the run tests, review criteria and a stopping condition before it starts.

Choose Opus, Sonnet, Grok or ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition. when the policy, the price or the task shape points elsewhere. A high benchmark score doesn't make a model the safest or most efficient default for a team. Fable fits hard work that benefits from sustained intent and justifies both the retention exception and the published premium.