Anthropic cuts Fable costs, loosens restrictions

Anthropic has released Fable 5.1, a new version of its AI model that cuts costs and eases content restrictions.

The update addresses two persistent pain points for developers: high token costs and overzealous content filters. Fable, part of Anthropic’s push into text-generation tools, has faced criticism for blocking harmless prompts or charging too much for long outputs. The company now says it has fine-tuned the model to reduce false positives in its safety checks while lowering pricing, likely in response to competition from open-source alternatives and other AI labs offering cheaper, more permissive models.

Fable 5.1’s token prices are now more competitive, though exact figures weren’t disclosed. The model also promises fewer unnecessary rejections of benign requests, a common frustration for users building creative or business applications. Anthropic has not detailed the technical changes but suggests the improvements come from refined training data and adjusted moderation thresholds.

This move signals Anthropic’s effort to balance safety with usability as it competes in a crowded AI market. If the changes work as intended, developers may adopt Fable more widely for tasks like storytelling, code generation, or customer support. The next test will be whether the model maintains its reliability while being more open—and whether users see real savings.