Claude Fable 5.2 is now generally available in GitHub Copilot

Developers on a Business or Enterprise Copilot plan cannot pick Claude Fable 5.2 today. The model landed in Copilot the same day Anthropic released it. The policy that turns it on sits in the admin's settings, off by default. On Pro+ and Max plans it is already available. The rollout covers Pro+, Max, Business and Enterprise accounts. The model picker will carry it across every Copilot surface. GitHub lists Visual Studio Code, Visual Studio and the Copilot CLI. It also lists the Copilot coding agent, the Copilot app, github.com and GitHub Mobile. The list ends with JetBrains IDEs, Xcode and Eclipse.

What GitHub says it is for

GitHub markets Fable 5.2 for "complex, multistep coding tasks across large codebases". GitHub says its own early testing showed Fable 5.2 solved more tasks than Fable 5.1. The model used fewer steps and tool calls to do so. It also finished hard tasks the earlier model missed. GitHub does not give a number for the difference. The score Anthropic reports is 49.1 percent on FrontierCode 2.0. That is up from 41.1 percent for Fable 5.1 on the same test.

The default is off for teams

A Business or Enterprise admin has to turn Fable 5.2 on. Nobody on the plan can pick it before that. GitHub says the policy is disabled by default. It lives in the Copilot settings for the organisation. On Pro+ and Max plans there is no admin step. The model is billed at provider list pricing under usage-based billing. GitHub does not publish a multiplier in the changelog.

A data retention step most Anthropic models do not have

GitHub says Fable 5.2 "requires data retention by default" for Anthropic's safety classifiers. Prompts and outputs are kept by Anthropic. GitHub says Anthropic does not use the kept data for model training. Eligible Enterprise customers can get zero data retention through the end of 2026. That is a time-bound exemption. GitHub says it is in place while Anthropic develops a feature it calls Enterprise Frontier Safeguards.

A Copilot admin at a shop that needs zero data retention should wait. Do not turn Fable 5.2 on for the team until the exemption is in place. A single developer on Pro+ or Max who picks Fable 5.2 agrees to the retention terms.

What is new in this model

Anthropic said Fable 5.2 generates output up to 30 percent faster than Fable 5.1. The standard price is the same. The company also opened a new fast mode. Fast mode trades a higher token price for up to 2.5 times the output speed per step. The standard price stays at 8 dollars per million input tokens and 40 dollars per million output tokens. GitHub's changelog does not mention fast mode. A Copilot user picking Fable 5.2 gets the standard price and the standard speed.

The vendor scores are Anthropic's own. Anthropic ran them on the same tests as Fable 5.1. On RoboDojo Sim, a robot simulation test, Fable 5.2 scored 64.9 percent against 27.6 for 5.1. On BenchCAD 2.1, a computer-aided design test, it scored 83.5 against 44.4 for 5.1.

What this is for a Copilot admin

A team moving from Fable 5.1 to Fable 5.2 pays the same list price per token. The vendor score on coding tasks is higher. The switch costs the admin step on Business and Enterprise. It also costs the kept-prompts term on every plan outside the exemption. That is the trade a Copilot admin reads this changelog to make.

Source

Claude Fable 5.2 in GitHub Copilot, GitHub Changelog.