What happened
GitHub has made Anthropic’s Claude Fable 5.1 generally available in GitHub Copilot. The model is offered to Copilot Pro+, Max, Business, and Enterprise users, although the rollout is gradual.
GitHub announced on September 1, 2026, that Claude Fable 5.1 from Anthropic is available in GitHub Copilot. GitHub identifies it as the latest model in Anthropic’s Mythos class and says it is designed for long-horizon, autonomous coding and knowledge-work tasks. The announcement specifically names deep codebase research, feature development, and complex agentic workflows as examples of the work for which the model is intended. These descriptions are GitHub’s claims about the model’s design and internal testing, not independently established performance results. The announcement therefore establishes availability and stated intended use, while leaving independent measurement outside the information provided here.
The model is available to Copilot Pro+, Max, Business, and Enterprise users. GitHub says users can select it from the model picker in Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot coding agent, the Copilot app on github.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode, and Eclipse. The rollout is gradual, so GitHub warns that the option may not immediately appear for every eligible user or organization. This means the general-availability announcement describes the product status while access can still vary during the rollout across the named users, organizations, and development environments.
Business and Enterprise administrators must enable the Claude Fable 5.1 policy in Copilot settings, and the policy is off by default. GitHub says enabling the policy acknowledges that the model requires data retention by default. When Claude Fable 5.1 is used under that default arrangement, Anthropic retains prompts and outputs to operate safety classifiers that detect harmful use. GitHub says the retained data is not used to train Anthropic’s models. The policy setting and retention acknowledgment are therefore part of the release conditions for the relevant plans, alongside the stated purpose of the retained prompts and outputs.
The release is billed at provider list pricing under usage-based billing, with further details directed to GitHub’s Copilot pricing information. GitHub also says that, unlike other Claude models available in Copilot, Claude Fable 5.1 requires retention by default. The source does not specify the model’s provider price, rate limits, usage caps, geographic availability, or a date by which the gradual rollout will be complete. Those omissions leave the announcement focused on availability, policy, billing method, and retention terms rather than a complete schedule or cost specification.
Why it matters
The release gives Copilot users access to a model GitHub describes as designed for long-running coding and knowledge-work tasks. It also introduces a significant data-governance condition: prompts and outputs are retained by default so Anthropic can operate safety classifiers.
The immediate significance is product access. Claude Fable 5.1 is being placed inside several of Copilot’s development surfaces, including an autonomous coding agent and command-line and IDE integrations. That makes the announcement relevant to organizations and individual developers already using Copilot who want to test a model aimed at work extending beyond short code suggestions. The source, however, does not establish how the model compares with other Copilot models in independent evaluations or in users’ production environments.
GitHub’s description places particular emphasis on tasks that can span substantial codebases and multiple stages of work. If the model performs as GitHub reports in its internal testing, its use could affect how Copilot users approach repository research, feature implementation, and more complex agentic workflows. The available evidence is limited: GitHub provides a qualitative account of internal testing but no test design, benchmark scores, sample sizes, failure rates, or comparisons with competing models.
The data-retention requirement is a material operational consideration. Organizations enabling the model must accept that prompts and outputs will be retained so Anthropic can run safety classifiers, even though GitHub says the retained data is not used for model training. This creates a distinction between model capability and deployment conditions: access depends not only on a plan and administrator approval, but also on an organization’s willingness to use the default retention arrangement or qualify for a separate exemption.
Eligible enterprise customers can use Claude Fable 5.1 and Claude Fable 5 with zero data retention through the end of the calendar year under a time-bound exemption. GitHub says the exemption will remain available while Anthropic rolls out Enterprise Frontier Safeguards, which are intended to provide automated safety monitoring and more options for data retention, including storage and encryption keys under the customer’s control. The source does not define the full technical or contractual scope of those safeguards.
What to watch next
The main issues to watch are rollout completion, enterprise approval and adoption, and the transition from the temporary zero-data-retention exemption to Anthropic’s Enterprise Frontier Safeguards. GitHub does not provide independent performance results or detailed availability timing in the source.
The first near-term question is how quickly the gradual rollout reaches each listed Copilot surface. GitHub says the model is generally available but also tells users to check back if they do not yet see it. The announcement does not state whether all listed integrations will receive access simultaneously, nor does it provide regional, account-level, or capacity-related availability details.
For Copilot Business and Enterprise customers, administrators will determine whether the model is enabled. GitHub says the policy is off by default, and approval for the zero-data-retention exemption does not automatically enable the model. Organizations must still have an administrator enable the Claude Fable 5.1 and, where applicable, Claude Fable 5 policies in Copilot settings. GitHub says account teams can determine eligibility, while GitHub Support cannot.
The most important longer-term checkpoint is the end of the calendar year. GitHub says continued zero-data-retention use of Fable 5 and Fable 5.1 will then require Enterprise Frontier Safeguards. The announcement does not say how many customers qualify for the exemption, what review standards apply, how the safeguards will operate in practice, or whether customers will be able to retain all current controls after the exemption ends.
Performance and reliability also remain open questions. GitHub reports strong internal testing on substantial, long-running coding tasks, but it does not publish detailed evidence in this announcement about accuracy, security, cost, latency, tool-use reliability, or the model’s behavior when tasks fail. Those missing details will matter for organizations deciding whether the model’s capabilities justify usage-based costs and the default data-retention arrangement.