SafetyPage

Lab system cards and frameworks · System card · Anthropic

Model Card Addendum: Claude 3.5 Haiku and Upgraded Claude 3.5 Sonnet

22 Oct 2024

Read at anthropic.com ↗Read at sourceAll rights reserved

Outline · 14 pages

  1. Introductionp. 1
  2. Knowledge Cutoffp. 1
  3. Evaluationsp. 1
  4. Computer Usep. 2
  5. Tool Use & Agentic Tasksp. 3
  6. Vision Capabilitiesp. 4
  7. Refusalsp. 5
  8. Reasoning, Coding, and Question Answeringp. 5
  9. Human Feedback Evaluationsp. 6
  10. Safetyp. 7
  11. Trust & Safetyp. 7
  12. Frontier Risk Evaluationsp. 9
  13. Ongoing Safety Commitmentp. 10
  14. Appendixp. 11
  15. Computer use examplesp. 11