Users on GitHub Copilot's higher individual and organizational plans can now choose a new model for complex coding tasks. Select Anthropic's Claude Fable 5.2 in the model picker. Enterprise and organization administrators must enable access first. Default use entails prompt and output retention.

1. A new model for complex coding tasks

GitHub announced the general availability of Claude Fable 5.2 in Copilot on November 9, 2026. It presented the model as an option for complex, multistep coding tasks involving large codebases.

GitHub's initial tests found that Fable 5.2 solved more tasks than Claude Fable 5.1 with fewer steps, tool calls, and requests to run external functions. The company said it also completed difficult tasks the previous model could not solve.

2. Check your plan and select it in the model picker

Eligible plans are Copilot Pro+ and Max for individuals, and Business and Enterprise for organizations. Standard Pro is excluded. The supported-model documentation lists the same plan coverage as earlier Fable models.

In development tools, users can select it in Visual Studio Code, Visual Studio, JetBrains IDEs, Xcode, and Eclipse. It is also available in the command-line tool Copilot CLI.

It is also available in the coding agent and Copilot app. Support extends to the website and GitHub Mobile, and GitHub is rolling it out gradually.

3. Extended features consume credits and add costs

Fable 5.2 supports a 1 million-token context, the input it can handle at once, for working with large codebases or multiple files together. Users can select extended context instead of the default size in Visual Studio Code and Copilot CLI.

Users can also adjust the reasoning level, which determines how deeply the model examines a problem before answering. This feature is available in those two environments and in Copilot cloud agent.

Providing more content or increasing reasoning depth brings benefits but raises token consumption and AI credit usage. GitHub recommends starting with the default context and reasoning level, reserving extended features for complex tasks.

According to GitHub's changelog, the model uses usage-based billing, charging for the amount used at the provider's listed rates. This is why users should check task-specific extended settings as well as model availability.

4. Review data retention before enabling organizational use

When Fable 5.2 is used in Copilot, Anthropic retains prompts and outputs to operate safety classifiers that identify harmful use. GitHub says this data is not used to train Anthropic's models.

Anthropic says it needs to analyze multiple requests together to identify abuse patterns such as repeated bypass attempts. It says temporarily collecting data helps detect patterns that individual requests do not reveal.

Other Claude models, excluding Fable 5, 5.1, and 5.2, operate with zero data retention. This means prompts and outputs are not stored.

Anthropic's separate policy states that platforms offering designated models retain prompts and outputs for 30 days, effective June 9, 2026. It covers existing zero-data-retention organizations not notified of exemption eligibility.

Copilot Business and Enterprise administrators must enable the model policy in settings. It is off by default, and enabling it acknowledges the retention requirements. Fable models are excluded from GitHub's data retention agreement. They are not automatically allowed by an organization's default model enablement policy either.

Approved enterprises can temporarily use zero data retention through the end of 2026. Eligibility checks and applications go through a GitHub account representative. Even after approval, an administrator must enable the model before use can begin.

Retained data is automatically deleted after 30 days, but Anthropic's policy makes exceptions for data flagged by safety systems or subject to legal retention duties. The zero-data-retention exemption is limited to internal operations, including developing and evaluating a company's own customer-facing products. Model endpoints or outputs cannot be provided externally.

Continued use after the end of 2026 requires Anthropic's Enterprise Frontier Safeguards. This framework will offer automated safety monitoring and an option for eligible customers to control storage and encryption keys directly.

Fable 5.2 expands Copilot's options for more complex coding tasks, but users must also weigh plan requirements, credit costs, and default data retention conditions.