Anthropic has officially released Claude Fable 5.1 and Claude Mythos 5.1, updating its flagship AI models to lower operational costs and address customer feedback regarding overzealous safety filters. Fable 5.1 delivers stronger performance while cutting standard costs by 25 percent and complex agentic tasks by up to 45 percent through reduced prompt caching prices.
The update introduces more precise safeguards that reduce false positives on harmless requests, such as basic biology questions. Claude Fable 5.1 is available across all major platforms, whereas Mythos 5.1 remains restricted to vetted participants in Project Glasswing.
Early enterprise testers reported strong performance improvements, particularly in coding speed and handling data ambiguities. Additionally, Anthropic announced Enterprise Frontier Safeguards, which will allow enterprise customers to store monitoring data on their own cloud servers.
Why it matters
Prompt caching price cuts reduce the cost of running autonomous, long-running agentic tasks by up to 45%.
Refined safeguards reduce false-positive blocks on benign biology and cybersecurity developer requests.
Enterprise Frontier Safeguards allow enterprise customers to retain monitoring data on their own cloud infrastructure.
Source: theverge.com
