Do Son 
Claude Fable 5 is back online. Developers using Claude Code can access the model through their existing subscription until July 7. However, there is a significant catch. Usage is capped at 50 percent of the subscription allowance. After July 7, Fable 5 will no longer be included in any subscription tier. From that point, developers must purchase additional credits to continue using the model. As a result, many developers are rushing to experience Fable 5 before the window closes.
How the Model Classifier Works
The Claude Code team has confirmed that a model classifier automatically evaluates the difficulty of each submitted task. Routine coding work and standard debugging may be flagged automatically. The system then downgrades those requests to a lighter model, such as Claude Opus 4.8. In practice, this creates a quiet problem. Developers who explicitly select Fable 5 may find their tasks reassigned to a different model without notice.
The Log Flag That Shocked Developers
One particularly striking detail has caught the attention of the wider developer community. A developer noticed that many of their tasks were being downgraded automatically. When they inspected the logs, they discovered a flag that read: TOO_DUMB_TO_NEED_FABLE.
In plain terms, the classifier had decided the task was too simple to justify using Fable 5. Rather than routing the request to the selected model, the system flagged it and downgraded it to Claude Opus 4.8. The label has since spread widely across social media, triggering considerable frustration.
Paying for Access Is No Guarantee
Developers point out a fundamental contradiction. They are already paying subscription fees to access Fable 5. Yet even within that paid allowance, the system can override their explicit model selection. What makes it worse is that the override comes with a dismissive internal label.
When the issue surfaced publicly, Claude Code admitted they had not expected developers to read the logs. This response only added fuel to the fire.
A Pattern of Corporate Tone That Refuses to Hide
Critics noted that Anthropic‘s tone in this incident was consistent with the company’s broader culture. Other companies caught using similarly contemptuous internal labels might deflect by blaming a rogue programmer. Anthropic, by contrast, let the Claude Code team respond directly. The company made no attempt to distance itself from the label.
Support Our Threat Intelligence
If you find our CVE report and cybersecurity news helpful, consider supporting our work.
Buy Me a Coffee
PayPal
