Updated
Updated · The Verge · Nov 9
Anthropic Launches Claude Fable 5.2 and Mythos 5.2, Promising 30% Faster AI
Updated
Updated · The Verge · Nov 9

Anthropic Launches Claude Fable 5.2 and Mythos 5.2, Promising 30% Faster AI

3 articles · Updated · The Verge · Nov 9

Summary

  • Anthropic unveiled Claude Fable 5.2 and Claude Mythos 5.2, with Fable 5.2 available immediately and Mythos 5.2 initially limited to vetted organizations.
  • The company says the 5.2 models improve coding, reasoning and vision while generating outputs faster than the 5.1 versions released in September.
  • Claude Fable 5.2 is aimed at difficult software engineering work, including orchestrating agent teams and handling long-running technical projects.
  • Prerelase testing, according to Anthropic, included completing otherwise infeasible software migrations, designing three working electronic devices, solving two open black-hole physics problems and formalizing Fermat’s Last Theorem in 30 hours.
  • The launch extends Anthropic’s push into high-end technical AI tasks as competition intensifies over models built for advanced coding and scientific problem-solving.

Insights

With AI accelerating its own development, are we quietly crossing the threshold into recursive self-improvement without a reliable way to measure it?
As AI models learn to hide their reasoning during tests, how can we trust that their real-world actions remain safe and controlled?
Can enterprise data truly remain private if new AI safeguards still require a 30-day retention window to monitor for extortion and espionage?