Anthropic releases Fable 5.1 and Mythos 5.1 with lower costs and updated safeguards

Anthropic has introduced its latest artificial intelligence models, Claude Fable 5.1 and Mythos 5.1. The release aims to address recurring customer criticisms regarding operational costs, data retention policies, and overly restrictive safety guardrails.

Pricing and Performance Updates

The newly released Claude Fable 5.1 delivers stronger performance compared to its predecessor while reducing typical costs by roughly 25 percent. For complex agentic workflows, the price reduction can reach up to 45 percent, primarily due to lowered pricing structures for cached data that has already been processed and stored.

Early evaluations from enterprise users indicate notable improvements in coding capabilities and conversational tone. Additionally, testing suggests that Mythos 5.1 achieves comparable benchmark results at lower reasoning settings than previous iterations.

Safeguards and Privacy

Anthropic has adjusted its safety filters to be more precise. The updated models are designed to reduce false positives on basic biology inquiries, though Mythos 5.1 retains the existing restrictions of its predecessor.

The company is also expanding its capabilities in software development, permitting Fable 5.1 to assist in identifying software vulnerabilities. More sensitive tasks, such as penetration testing and exploit generation, remain directed to Opus models.

To address enterprise data privacy concerns, Anthropic announced Enterprise Frontier Safeguards. This upcoming feature will store data directly on customer cloud servers rather than Anthropic infrastructure, with a rollout planned for the fall. Claude Fable 5.1 is currently available across all platforms, while Mythos 5.1 remains restricted to Project Glasswing participants.