Anthropic’s latest models undercut competitor pricing while introducing enterprise-grade zero-data-retention safeguards and more precise cybersecurity protections.
Anthropic announced Claude Fable 5.1 and Claude Mythos 5.1 on September 1, positioning them as “the world’s most advanced models for coding and knowledge work” . The two share the same base model but differ in safeguards: Fable 5.1 is generally available, while Mythos 5.1 requires approved access for high-risk domains like biology and cybersecurity .
The headline improvement is cost reduction. By cutting cache-read pricing by 75% to $0.25 per million tokens, Anthropic estimates Fable 5.1 will cost approximately 25% less than Fable 5 for typical workloads, with savings up to 45% for highly agentic tasks .
Anthropic also introduced Enterprise Frontier Safeguards (EFS) , allowing customers to run models on their own infrastructure with zero data retention — addressing enterprise privacy concerns that previously limited Fable adoption . Additionally, Fable 5.1 can now be used for vulnerability discovery while still blocking exploit generation, a nuanced shift in cybersecurity policy .
The model launch comes amid reports that Anthropic is targeting an October IPO at a valuation exceeding $2 trillion .
