Claude Fable 5.2 arrives in GitHub Copilot at $8/$40 with data retention

Claude Fable 5.2, the model Anthropic released on November 9, 2026, became generally available in GitHub Copilot on the same day. GitHub's changelog lists it for Copilot Pro+, Max, Business and Enterprise, billed at Anthropic's list price under Copilot's usage-based billing: $8 per 1M input tokens and $40 per 1M output tokens, which is 20% below the $10 and $50 that Copilot charges for Claude Fable 5.1. Copilot Free and Copilot Pro are not on the list of plans.

The model reaches almost every Copilot surface at once, from Visual Studio Code, Visual Studio, JetBrains IDEs, Xcode and Eclipse to the Copilot CLI, the Copilot coding agent, the Copilot app, github.com and GitHub Mobile on iOS and Android. GitHub describes the rollout as gradual, so the model can take some time to appear in the model picker for every account, and it positions Fable 5.2 for complex, multistep coding work across large codebases, saying that in early testing it solved more tasks than Fable 5.1 with fewer steps and tool calls.

The condition attached to the launch concerns data. Fable 5.2 requires data retention by default: Anthropic keeps prompts and outputs to run safety classifiers that detect harmful use, and GitHub states that this retained data is not used to train Anthropic's models. For Copilot Business and Enterprise, the model stays off until an administrator enables the Claude Fable 5.2 policy, and turning that policy on counts as accepting the retention requirement. Prices in this report are as listed on GitHub's official models and pricing page and Anthropic's official model page on November 10, 2026.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣

CLAUDE FABLE 5.2 IN GITHUB COPILOT: PLANS, SURFACES AND ROLLOUT.

GitHub has kept the plan scope it used for the previous Fable release. Claude Fable 5.1 reached Copilot on September 1, 2026 for the same four plans, billed at provider list price and disabled by default for organizations, three months after Claude Fable 5 became generally available in Copilot in June 2026. Fable 5.2 follows the same pattern, so a developer on the $10 Copilot Pro plan who wants Anthropic's top Fable model in the editor still has to move to Pro+ or Max.

The table below combines the plan prices and monthly AI credit allowances published on GitHub's plans page with the availability stated in the Fable 5.2 changelog entry. One AI credit equals $0.01.

........

Copilot plan	Monthly price	Included AI credits	Claude Fable 5.2
Copilot Free	$0	Allowance, auto model selection only	Not available
Copilot Pro	$10	1,500 (1,000 base + 500 flex)	Not available
Copilot Pro+	$39	7,000 (3,900 base + 3,100 flex)	Available
Copilot Max	$100	20,000 (10,000 base + 10,000 flex)	Available
Copilot Business	$19 per seat	1,900 per user, pooled	Available after admin enables policy
Copilot Enterprise	$39 per seat	3,900 per user, pooled	Available after admin enables policy

........

Individual subscribers on Pro+ and Max select the model directly in the picker once the rollout reaches their account. In organizations, the policy gate applies to every surface at the same time, so a developer who cannot find Fable 5.2 in VS Code or on github.com should first check whether the organization has turned on the policy before assuming the gradual rollout is the cause.

··········

CLAUDE FABLE 5.2 PRICING IN GITHUB AI CREDITS.

Copilot converts every request into AI credits from per-token prices. Input, cached input, cache writes and output are priced separately for each model, and the total is turned into credits at 100 credits per dollar. GitHub's models and pricing page lists Fable 5.2 with the same input and output prices that Anthropic publishes for its own API, so the only change from Fable 5.1 is a straight 20% cut on each line.

........

Model in Copilot	Input	Cached input	Cache write	Output	Output in AI credits
Claude Fable 5	$10.00	$1.00	$12.50	$50.00	5,000
Claude Fable 5.1	$10.00	$0.25	$12.50	$50.00	5,000
Claude Fable 5.2	$8.00	$0.20	$10.00	$40.00	4,000
Claude Opus 5.5	$4.00	$0.20	$5.00	$20.00	2,000
Claude Sonnet 5.5	$2.00	$0.10	$2.50	$10.00	1,000

........

All prices are USD per 1M tokens. GitHub lists a single default pricing tier for the three Fable models, with no separate long-context rate, although the model itself has a 1M-token context window and up to 128K output tokens according to Anthropic. GitHub's supported-models page notes more generally that larger context and higher reasoning settings consume more AI credits, because they generate more tokens.

The cut narrows the gap with Claude Opus 5.5 without closing it. Fable 5.2 still costs twice as much as Opus 5.5 per token and four times as much as Sonnet 5.5, so teams that already route routine edits to cheaper models have no pricing reason to change that habit. The saving is largest on the long agent tasks where teams were already choosing Fable for its capability.

··········

WHAT A CLAUDE FABLE 5.2 CODING SESSION COSTS ON COPILOT PLANS.

To make the allowances concrete, the example below models one long agent session on a large repository. The assumptions are illustrative: 200,000 uncached input tokens, 1.2 million cached input tokens read back as the agent revisits the same files, 100,000 tokens written to cache and 60,000 output tokens. Prices are GitHub's official Copilot rates listed above.

With Fable 5.1, that session costs $2.00 for uncached input, $0.30 for cache reads, $1.25 for cache writes and $3.00 for output, a total of $6.55, or 655 AI credits. The same session on Fable 5.2 costs $1.60, $0.24, $1.00 and $2.40, a total of $5.24, or 524 credits. On Opus 5.5 the total is $2.74 (274 credits), and on Sonnet 5.5 it is $1.37 (137 credits).

........

Allowance	Fable 5.1 sessions	Fable 5.2 sessions	Opus 5.5 sessions
Copilot Pro+ (7,000 credits)	10	13	25
Copilot Max (20,000 credits)	30	38	72
Business, per user (1,900 credits)	2	3	6
Enterprise, per user (3,900 credits)	5	7	14

........

The session counts are whole sessions that fit within the included credits, before any additional paid usage. For a company with 50 Copilot Enterprise seats, the shared monthly pool is 195,000 credits, enough for about 372 sessions of this size on Fable 5.2 against about 297 on Fable 5.1. Pooled credits are shared at the billing entity level, so a handful of heavy Fable users can consume a large share of a team's allowance, which is why the budget controls described below deserve attention before the policy is switched on.

··········

[[ADPLUS_300x250_1]]

DATA RETENTION RULES FOR CLAUDE FABLE 5.2 IN COPILOT BUSINESS AND ENTERPRISE.

Fable 5.2 inherits the retention regime that GitHub applied to Fable 5 and Fable 5.1. Other Claude models in Copilot, with the exception of the three Fable releases, run under zero data retention (ZDR). The Fable models sit outside GitHub's default model availability policy because they are not covered by GitHub's data retention agreement, which is why they are disabled by default for organizations.

Anthropic's own model page for Fable 5.2 states that the model requires 30-day data retention on Anthropic's platform and is not available under zero data retention unless Anthropic expressly authorizes it. GitHub's changelog does not state a retention period for requests made through Copilot; it says that Anthropic retains prompts and outputs to run safety classifiers and does not use them for training.

There is a temporary route for companies that cannot accept retention. Approved enterprises can use Fable 5.2, Fable 5.1 and Fable 5 with zero data retention until the end of the 2026 calendar year under a time-bound exemption, with requests routed to a ZDR endpoint. After that date, continued use requires Anthropic's Enterprise Frontier Safeguards (EFS), which adds automated safety monitoring and retention options such as customer-controlled storage and encryption keys. Eligibility goes through the customer's GitHub account team; GitHub Support cannot grant it, and approval does not switch the model on, because an administrator still has to enable the policy. GitHub's documentation also limits use under these terms to internal operations, without making the models or their outputs available externally.

··········

HOW ADMINISTRATORS ENABLE CLAUDE FABLE 5.2 AND CONTROL SPEND.

For Copilot Business and Enterprise, enabling the model is a policy change in Copilot settings. The Claude Fable 5.2 policy is a separate switch from the Fable 5 and Fable 5.1 policies, so an organization that approved Fable 5.1 in September does not get Fable 5.2 automatically. Before turning it on, legal and security teams need to confirm that the retention requirement is acceptable for the repositories developers will use with it, or that the company has obtained the ZDR exemption through its account team.

Spend control follows the usage-based billing rules GitHub applies to all models. Included credits reset at 00:00 UTC on the first day of each month and do not carry over. When the pool runs out, additional usage at $0.01 per credit is enabled by default for organizations and enterprises, and it stops only if an administrator disables the "AI credits paid usage" policy in AI Controls. User-level budgets can cap a single developer's consumption, with a $0 budget blocking that user at once, and cost-center, organization and enterprise budgets cap metered charges after the pool is exhausted. GitHub's billing documentation states that there is no automatic fallback to cheaper models when a budget runs out, and it does not describe a per-model spending cap.

Code completions and next edit suggestions are not billed in credits and stay unlimited on paid plans, so the Fable 5.2 cost applies to chat, agent and coding agent requests that select the model. For a wider view of how Copilot's credit pricing compares with Cursor, Claude Code and Codex, see our guide to AI coding agent pricing. Enterprises that already received Copilot through GitHub Enterprise should also read how Copilot access with no seat charge bills AI usage from the first request, since that option includes no credit allowance at all.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁤

CLAUDE FABLE 5.2 BEHAVIOR NOTES FOR COPILOT USERS.

Anthropic's documentation for Fable 5.2 describes several behaviors that developers will notice inside an editor. Thinking is adaptive and always on, its depth is set by an effort parameter with a default of high, and at high effort and above the model reasons before almost every reply. Anthropic also notes that on long multistep tasks turns can end early, more often at low and medium effort, that long runs of tool calls may produce little user-facing text, and that without explicit guidance on ownership the model tends to extend existing structure and report nearby problems rather than fix them.

These notes describe the model on Anthropic's platform; GitHub has not published which effort settings Copilot uses for Fable 5.2 on each surface. Developers who call the model directly through the Claude API face a separate set of breaking changes from Fable 5.1, including rejected forced tool use and the removal of manual thinking budgets, which our report on the Claude Fable 5.2 launch and pricing covers alongside the new API prices. Inside Copilot those request-level details are handled by GitHub, so switching models in the picker requires no code change.

··········

CLAUDE FABLE 5.2 IN GITHUB COPILOT: CONFIRMED DETAILS AND OPEN POINTS.

The confirmed facts are specific. Claude Fable 5.2 is generally available in GitHub Copilot from November 9, 2026 on Pro+, Max, Business and Enterprise, across the IDEs, the CLI, the coding agent, the Copilot app, github.com and GitHub Mobile. It costs $8 per 1M input tokens, $0.20 for cached input, $10 for cache writes and $40 for output, 20% less than Fable 5.1 on every line, and in the session modeled above that cut lifts a Pro+ allowance from 10 to 13 long Fable sessions a month.

Several points remain open. GitHub has not stated the retention period for Fable requests made through Copilot, the effort settings Copilot applies, or whether Copilot Pro will ever receive Fable models. The ZDR exemption ends with 2026, and the terms of Enterprise Frontier Safeguards inside Copilot have not been published, so organizations that rely on the exemption have about seven weeks to decide whether Fable models stay in their approved list. GitHub's supported-models reference did not yet list Fable 5.2 when checked on November 10, 2026, while the pricing page already did.

··········

FOLLOW US FOR MORE.

·····

DATA STUDIOS

·····

datastudios.org