GitHub Copilot: Claude Fable 5 Now Generally Available

GitHub Copilot Gains Claude Fable 5, Anthropic's Most Powerful Model

On June 9, 2026, GitHub Copilot added Claude Fable 5 to its model roster, marking the arrival of Anthropic's first Mythos-class model in a mainstream coding assistant. Fable 5 represents a significant capability leap from previous Claude generations, and its inclusion in Copilot gives Pro+, Max, Business, and Enterprise subscribers access to one of the highest-performing models on agentic coding benchmarks available today.

What Makes Fable 5 Different

Fable 5 is the first model in Anthropic's new Mythos class, a tier built specifically for long-horizon, autonomous tasks that unfold over extended sessions. Rather than excelling in single-turn exchanges, the model is optimized for sustained multi-step work: large codebase migrations, complex refactors, and knowledge-intensive engineering tasks that require holding significant context across many tool calls.

On SWE-Bench Pro, the leading benchmark for autonomous software engineering, Fable 5 scores 80.3%, compared to 69.2% for Opus 4.8 and 58.6% for GPT-5.5. On FrontierCode, a benchmark for harder coding challenges, it reaches 29.3% versus Opus 4.8 at 13.4%. GitHub's internal benchmarks on autonomous coding workflows found that Fable 5 completed equivalent work with fewer tool calls and lower token consumption than Opus-tier models, which has a direct impact on AI credit costs for users.

Where It Works in Copilot

Once enabled, Claude Fable 5 is selectable via the model picker across the full range of Copilot surfaces: VS Code (in chat, ask, edit, and agent modes), Visual Studio, the Copilot CLI, GitHub.com, JetBrains, Xcode, Eclipse, and mobile apps. The rollout is gradual, so availability may vary slightly in the first days after launch.

The Data Retention Requirement

Claude Fable 5 comes with a data retention requirement that distinguishes it from every other Claude model currently available in Copilot. Anthropic retains prompts and outputs for up to 30 days in order to operate safety classifiers designed to detect harmful or abusive use. After 30 days, the data is deleted and is not used for model training. This applies only to Fable 5; Claude Opus 4.8, Sonnet 4.5, and Haiku 4.5 maintain Zero Data Retention (ZDR) as before.

Developers and organizations for whom ZDR is a firm requirement will need to weigh this trade-off carefully before opting into Fable 5.

Admin Requirements and Pricing

For Business and Enterprise plans, admins must explicitly enable the Claude Fable 5 model policy in Copilot settings before developers can select it. The model defaults to off at the org level. On Pro+ and Max plans for individuals, it is available directly through the model picker.

Fable 5 is priced at $10 per million input tokens and $50 per million output tokens under usage-based billing. As an introductory offer, it is included at no additional cost through June 22, 2026, for Pro, Max, Team, and seat-based Enterprise plans. After that date, usage moves to the standard AI credit pricing.