Hasty Briefsbeta

Bilingual

Claude Fable 5

4 hours ago
  • #Cybersecurity
  • #Claude Fable 5
  • #AI Models
  • Anthropic launched Claude Fable 5, a state-of-the-art AI model with safeguards for general use, and Claude Mythos 5 for select cyberdefenders with some safeguards lifted.
  • Fable 5 is a Mythos-class model that excels in software engineering, knowledge work, vision, scientific research, and long/complex tasks, priced at $10/million input tokens and $50/million output tokens.
  • Safeguards include classifiers that detect and redirect cybersecurity, biology/chemistry, and distillation queries to Claude Opus 4.8, with less than 5% of sessions triggering fallback.
  • Fable 5 demonstrates strong performance in code migration, autonomous operation, vision tasks, memory, and drug design, with early feedback highlighting its superior reasoning and agentic capabilities.
  • Mythos 5, offered through Project Glasswing and a trusted access program, matches Fable 5's capabilities but removes cybersecurity safeguards for authorized users and will expand to biology researchers.
  • Safety measures include robust classifiers tested against jailbreaks, a new 30-day data retention policy for security, and plans to refine safeguards to reduce false positives over time.