LLMMarch 4, 2024Anthropic

The Claude 3 Model Family: Opus, Sonnet, and Haiku

Anthropic

Abstract

We introduce the Claude 3 family of AI models: Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku. These models represent a significant leap in capabilities across reasoning, math, coding, multilingual understanding, and vision. Claude 3 Opus achieves near-human-level performance on expert knowledge benchmarks and sets new standards for AI safety and ethical behavior.

Key Findings

  • 1Introduced three tiers (Opus, Sonnet, Haiku) for different capability-cost tradeoffs
  • 2Claude 3 Opus achieved near-human performance on graduate-level expert tasks
  • 3Added native vision capabilities across all model tiers
  • 4Demonstrated industry-leading safety and reduced hallucination rates
  • 5Showed strong multilingual capabilities across major world languages

Impact & Significance

Claude 3 established Anthropic as a frontier AI lab with models competitive with GPT-4 and Gemini Ultra. The tiered approach influenced how AI companies structure their model offerings for different use cases.

Related Tools

Read Full Paper