Fable is now available on Cyrus

Anthropic announced Claude Fable 5 and Claude Mythos 5 on June 9. Fable, a Mythos-class model, is now available on Cyrus as a Claude model option.

That means your team can configure Fable for Claude sessions without changing the rest of your Cyrus workflow. It is available when you want to try a model tuned for long-horizon coding and agent work, while keeping Opus and Sonnet available in the fallback chain when needed.

Why Fable matters for agent work

Most Cyrus work happens over many steps: reading an issue, inspecting a codebase, planning a change, editing files, running checks, responding to feedback, and opening a PR. Long-running sessions need a model that can stay oriented across that whole loop.

Fable is designed for that kind of sustained agent work. In Cyrus, that makes it a good option for complex implementation tasks, refactors, and issues where the agent needs to hold a plan while adapting to what it learns from the repository.

We are keeping the claims here intentionally grounded. Anthropic's announcement goes deeper on the model family, safety design, and evaluation results. If you want the model-level details, read Anthropic's Fable 5 announcement.

How to use Fable in Cyrus

Teams can choose Fable as their default Claude model from Behaviours. Once saved, new Claude sessions for that team can run on Fable by default.

You can also try Fable on a single Linear issue before changing team defaults. Add the fable label to the issue before assigning it to Cyrus, and Cyrus will run that Claude session on claude-fable-5.

Fable support is available in hosted Cyrus, and self-hosted teams can update to the latest Cyrus release to pick up the fable model selector. When Fable is selected, Cyrus uses the configured fallback path to Opus and then Sonnet if needed.

If you're not yet using Cyrus, get started today and try Fable on your next long-running development task.