Anthropic released two flagship AI models (Fable & Mythos 5.1) with significant cost and privacy improvements. Prompt caching prices cut 75%, reducing total operational costs for complex workloads. New Zero Data Retention and Enterprise Frontier Safeguards features let organizations run models on their own infrastructure with customizable safety controls, while guaranteeing enterprise data won't be used for model training without explicit permission—a direct response to SME and enterprise concerns about data security and TCO.