A report suggests that Claude Code’s Fable model selection may now save a rolling choice rather than pinning users to one specific model version. That could make it possible for a saved Fable selection to route to a newer model, but it does not confirm that Fable 5.5 is being tested or has been released.

In a post on X, @LuminaBench says the Claude Code /model picker previously saved the current Fable version’s ID, while the newer behavior follows “the newest Fable” in the same way the poster says Opus and Sonnet did during earlier testing. The reported change could affect which model is served when someone returns to a saved Fable selection.

What the reported Claude Code change does

A pinned model selection identifies a particular version. A rolling selection instead preserves the model family—Fable in this case—and can potentially follow whichever version is treated as the newest available model in that family.

That distinction matters because a user who saved Fable may not necessarily remain tied to the exact Fable version that was available when the choice was created. It remains unclear whether every Fable selection behaves this way, which model a user would receive, or how broadly the routing is available.

Screenshot text describes a saved Fable choice following the newest model like Opus and Sonnet.
Screenshot text describes a saved Fable choice following the newest model like Opus and Sonnet.

Image credit: @LuminaBench on X

Why the behavior is linked to possible Fable 5.5 testing

The report connects the Fable behavior to earlier observations involving Opus and Sonnet. The poster says those model families showed similar routing behavior before variants described as 5.5 were released, and interprets the matching Fable behavior as a possible sign of greyscale testing—a limited or controlled rollout to some users or requests before a wider release.

That interpretation is an analogy based on an observed pattern, not proof of a new Fable version. A rolling model choice can show that newer routing may be possible without showing why it was enabled or identifying the model behind a particular response.

The thread also mentions occasional strong outputs and apparent knowledge beyond a reported cutoff. Those are anecdotal observations from the post, not benchmark results or independently reproducible evidence of Fable 5.5’s capabilities.

What this does not confirm

The report does not establish that:

  • Fable 5.5 exists as a publicly announced model;

  • Anthropic is currently testing Fable 5.5;

  • Fable 5.5 has been released or is broadly available in Claude Code;

  • every saved Fable choice routes to a newer model;

  • the behavior reflects a deliberate greyscale rollout;

  • the model has particular capabilities, benchmarks or knowledge limits; or

  • there is a release date, access policy, price or availability scope.

The supplied image captures LuminaBench’s description of the reported model-picker behavior. Its text does not identify the model being served or establish why the routing was enabled.

How to interpret the report

The strongest supported conclusion is that the described Claude Code behavior may allow a saved Fable selection to follow a newer Fable model instead of remaining fixed to one version. That makes newer routing possible, but it leaves the identity, rollout status and timing of any underlying model unresolved.

For now, “Fable 5.5 Claude Code” is best treated as a possibility raised by an unverified routing observation—not as evidence of a confirmed release or public test.

Source: Source on X

Sources