Anthropic released Claude Fable 5, its most capable model to date, earlier this week. According to the company, Fable 5 outperforms all previous models on nearly every benchmark—particularly in software engineering, knowledge work, and scientific research. It was priced at ten dollars per million input tokens and fifty dollars per million output tokens, double the cost of Claude Opus. The launch was short-lived. Within three days, the U.S. Government ordered Anthropic to pull the model offline after a critical jailbreak vulnerability was discovered. Anthropic expects it to return within two weeks. Meanwhile, the AI research community has been scrutinizing Fable 5's system prompt, with critics arguing that Anthropic's approach relies too heavily on prescriptive safety rules that don't generalize well to novel situations.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton