Claude Fable 5 Now Generally Available in GitHub Copilot
Key Takeaways
- ▸Claude Fable 5 is Anthropic's first Mythos-class model, optimized for autonomous coding and long-horizon knowledge-work tasks
- ▸The model delivers improved performance with fewer tool calls and lower token consumption than previous Opus-tier models
- ▸Claude Fable 5 requires 30-day data retention for safety classifiers—a unique requirement among Claude models in GitHub Copilot
Summary
Anthropic has announced the general availability of Claude Fable 5 in GitHub Copilot, the first model in its new Mythos class designed for long-horizon, autonomous coding and knowledge-work tasks. The model is now accessible to all Copilot Pro+, Max, Business, and Enterprise users through Visual Studio Code, the GitHub Copilot cloud agent, and GitHub Mobile, with a gradual rollout beginning immediately.
In internal benchmarks on autonomous coding workflows, Claude Fable 5 demonstrates improved efficiency compared to previous Opus-tier models, completing equivalent work with fewer tool calls and lower token consumption. The model is billed at standard provider list pricing under GitHub Copilot's Usage Based Billing structure.
A notable distinction for Claude Fable 5 is its data retention requirement for safety operations. Unlike other Claude models in GitHub Copilot (Opus 4.8, Sonnet 4.5, and Haiku 4.5), Fable 5 requires Anthropic to retain prompts and outputs for up to 30 days to operate its safety classifiers that detect harmful or abusive use. Data is deleted after 30 days and is not used for model training. For Copilot Enterprise and Business administrators, the Claude Fable 5 policy is disabled by default and must be manually enabled to make the model available to their organizations.
- Available to all Copilot subscription tiers with gradual rollout; administrators must explicitly enable the policy for Business and Enterprise use
- Billed at provider list pricing under GitHub Copilot's existing Usage Based Billing model
Editorial Opinion
Claude Fable 5 represents a meaningful expansion of Anthropic's model portfolio in GitHub Copilot and signals the company's confidence in deploying specialized architectures for enterprise coding workflows. The data retention requirement, while transparent and limited, introduces a privacy-conscious tradeoff that organizations will need to evaluate—particularly for sensitive codebases. For developers seeking autonomous coding capabilities with stronger performance metrics, Fable 5 offers a compelling option, though teams handling proprietary code may prefer the zero-data-retention alternatives. This launch reinforces the competitive pressure on GitHub Copilot and Microsoft to support multiple state-of-the-art models rather than defaulting to OpenAI's offerings.
