Claude Fable 5.1 in Cursor: Pricing and Uses

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. is the model Cursor now lists in place of Fable 5, built for long-running, complex and asynchronous work. Cursor describes three strengths: cleaner communication than Fable 5 on long agent sessions, autonomous multi-step work that holds intent across a long session and drives tasks to completion with fewer check-ins, and deep reasoning, with a thinking variant for problems that reward extra deliberation.

Input and output rates match Fable 5, and the one price change is on cache reads.

The strengths list is Cursor's description of the model, and the only comparison it makes is against Fable 5.

Fable 5.1 bills at $10 per million input tokens and $50 per million output tokens, identical to Fable 5. Prompt-cache reads are $0.25 per million tokens, 75% below Fable 5's $1.00 cache-read rate, and cache writes stay at $12.50 per million tokens.

Fable 5.1 is a third-party model, so it is not part of the usage included with Cursor plans. On 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. it is billed at these rates from your prepaid credit. On legacy Cursor plans it draws from 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.

Cursor's models and pricing page adds that third-party models are available on Premium and above at the model's API list price from the first request, paid from prepaid credit, and that Plus and Free do not include them. Top-ups add the credit, Auto Top-up refills it below a threshold you set, and a monthly additional usage limit caps paid additional usage each month. Fable 5.1 also supports a thinking variant for deeper reasoning, and Cursor recommends the high thinking variant for the strongest results.

The docs do not give a figure for how much the cache-read cut saves in practice.

Cursor also states the cost relationship to a cheaper model. Input and output rates are about 2.5x Claude Opus 5.5Anthropic's top Opus-tier model in Cursor at $4 per million input tokens and $20 per million output, 20% below Claude Opus 5, which it replaces. Press Enter for the full definition., which Cursor says now scores higher on CursorBenchCursor's internal benchmark that scores models on both task performance and token efficiency, not accuracy alone. Press Enter for the full definition.. In other words, the more expensive model is not the higher-scoring one on that benchmark.

The 2.5x figure is consistent with the listed rates. Fable 5.1 is $10 input and $50 output per million tokens, and Opus 5.5 is $4 input and $20 output, so each ratio works out to 2.5. Cache reads differ by less: $0.25 for Fable 5.1 against $0.20 for Opus 5.5.

Fable 5.1 is on by default for individual customers and Team accounts that have 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. disabled. It is off by default for every Team with Privacy Mode enabled and for every Enterprise customer, regardless of Privacy Mode settings, until an admin explicitly enables it.

The reason for the opt-in is Anthropic's data handling. When Fable 5.1 is used, Anthropic stores agent input and output data to support automated and human harm-prevention processes, and that holds regardless of 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. settings. Cursor says this data is not used to train or improve Anthropic's models or products.

Enabling and using the model does not change your 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. settings or Cursor's own retention of your data.

To enable it, the docs point to the restricted-models page in the Cursor Dashboard, where the Data Retention Policy can be viewed and approved for your Team, Organization or account. For Teams accounts, approving the model policy applies to the entire Team.

1. 1Open the Fable 5.1 page in the Cursor Dashboard (restricted models) and read the Data Retention Policy.
2. 2Approve the policy. On a Team account this applies to the entire Team.
3. 3On Enterprise, approve model availability at the Organization level.
4. 4Enterprise admins can restrict which user groups can use the model with Model Access controls.

On Enterprise the docs say availability is approved at the Organization level but continues to be controlled at the team and group level through Model Access controls.

Cursor summarizes Anthropic's policy for Fable models in four statements and links the full document. The four are in the table below.

The first row carries two exceptions: a safety investigation or a legal hold can keep data past 30 days, so 30 days is the default period.

Fable 5.1 runs with tight security guardrails. When a request trips one of them, Cursor routes it to Claude Opus automatically, so your work continues without an error and you do not need to retry or switch models yourself.

In other words, a request that trips a guardrail is answered by Claude Opus instead.

Fable 5.1 has access to all agent tools when used with Cursor, so the fallback changes the model behind the tools, not what the agent can do.

Cursor's own page lists two limitations, and they are the best guide to when it earns its price. The first is the data-retention opt-in and the admin approval it needs. The second is cost: rates are about 2.5x Claude Opus 5.5Anthropic's top Opus-tier model in Cursor at $4 per million input tokens and $20 per million output, 20% below Claude Opus 5, which it replaces. Press Enter for the full definition., which now scores higher on CursorBenchCursor's internal benchmark that scores models on both task performance and token efficiency, not accuracy alone. Press Enter for the full definition..

So the case for Fable 5.1 is the shape of the work rather than a general ranking. Long-running, autonomous multi-step work and problems that reward extra deliberation are what the docs point at.

Cursor recommends the high thinking variant for the strongest results.

The policy is a team-wide opt-in, and on 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. and Enterprise accounts the model is unavailable until an admin explicitly enables it.