This article is a technical commentary and implementation example created using AI. The included code and procedures are based on primary sources, but the author has not verified their operation on actual hardware. Behavior may vary depending on the environment and version.
According to information published in the GitHub Changelog, Anthropic’s “Claude Fable 5.1” is now generally available (GA) in GitHub Copilot. Belonging to Anthropic’s Mythos class, this latest model is designed to efficiently handle long-running autonomous coding and advanced knowledge work tasks. Based on the official announcement, this article outlines the supported environments, prerequisites for use, and data retention specifications.
Overview and Key Features of Claude Fable 5.1
According to primary sources, Claude Fable 5.1 is introduced as a model optimized for long-term, complex coding tasks.
Long-Running Autonomous Task Processing: Internal tests have shown that it delivers superior performance in large-scale codebase investigations, new feature development, and complex agentic workflows.
Mythos Class Positioning: Provided as Anthropic’s latest generation model, it comes with different operational policies and safety feature requirements compared to traditional models.
Supported Client Environments
Claude Fable 5.1 is being rolled out via a staged rollout format. Therefore, the primary source notes that some environments may not display it on the screen just yet. The main available interfaces and IDEs include the following:
Visual Studio Code
Visual Studio
Copilot CLI
GitHub Copilot coding agents
GitHub Copilot apps
github.com
GitHub Mobile (iOS and Android)
JetBrains IDEs
Xcode
Eclipse
To use the model, select it from the model picker in each supported environment. Regarding pricing, usage-based billing applies based on the provider’s list price.
Administrator Policy Enforcement and Data Retention Requirements
To use Claude Fable 5.1 within an organization, a Copilot Enterprise or Copilot Business plan administrator must explicitly enable the dedicated policy from the Copilot settings screen. This policy is set to off by default, and users within the organization cannot use the model while it remains disabled.
Default Data Retention Specifications
Unlike many other Claude models, Claude Fable 5.1 requires data retention, including prompts and outputs, by default in order to run safety classifiers.
Purpose of Data Retention: Used exclusively to operate safety classifiers for detecting harmful use.
Non-Use for Model Training: It is explained that the retained data is not used for Anthropic’s model training.
Implications of Enabling the Policy: Enabling the Claude Fable 5.1 policy is considered acknowledgment and acceptance of this data retention requirement.
Zero Data Retention (ZDR) Access for Eligible Enterprises
Zero Data Retention (ZDR) access is also available for certain eligible enterprise customers.
Time-Bound Exemption: Until the end of the calendar year, Claude Fable 5.1 and Fable 5 are available with ZDR as a time-bound exemption accompanying the deployment of Anthropic’s “Enterprise Frontier Safeguards (EFS).”
Future Requirements: After the period ends, the introduction of EFS—which allows companies to manage their own storage and encryption keys while enabling automated safety monitoring—will become a condition for continued use.
Checking Eligibility: If you wish to confirm whether your organization is included, you must contact your GitHub account team rather than GitHub Support.
Enabling Steps: Even when approved, administrators must still manually enable the policy from the Copilot settings.
Conclusion
Claude Fable 5.1 has been added to GitHub Copilot as a powerful option to support long-running autonomous coding and advanced agentic workflows. At the same time, there are operational considerations that differ from conventional models, such as data retention specifications driven by the operation of safety classifiers and the requirement for administrators to explicitly enable the policy. When considering adoption, it is important to review official documentation and account team guidance in light of your organization’s security policies and data requirements.
コメント