Claude Fable 5: The first Mythos model is powerful, expensive, and heavily filtered

Editor's take

Anthropic's Claude Fable 5, the inaugural model from its advanced Mythos tier, demonstrates top-tier performance, notably achieving 95% on SWE-bench Verified, yet arrives with a significantly higher price point, costing 10x or 50x more per million tokens than its predecessor, Opus 4.8.

This release signals a strategic shift for Anthropic, prioritizing raw capability and potentially targeting enterprise applications where cost is a secondary concern to advanced reasoning and coding prowess. The substantial price increase, coupled with stringent safety filters, suggests a focus on controlled, high-value deployments rather than broad consumer accessibility, positioning Fable 5 as a premium offering in the increasingly competitive LLM market dominated by players like OpenAI and Google.

Future developments will reveal if this premium pricing model proves sustainable and if Fable 5's performance justifies the substantial cost for developers and businesses. Scrutiny will also focus on the efficacy and potential limitations imposed by its "heavily filtered" nature, particularly in nuanced or creative use cases, and whether Anthropic can maintain its safety edge without sacrificing utility.