Claude Models and Versions

Anthropic offers a spectrum of models with different capabilities, balancing power, speed, and cost. Each version represents significant advancements in AI capabilities.

O

Opus

"Magnum opus" - largest work

  • Highest intelligence capability
  • Most thorough analysis
  • Higher cost, slower inference
  • Best for complex tasks
S

Sonnet

Medium-sized poem

  • Balance of speed and capability
  • Moderate cost
  • Great for everyday use
  • Most popular model
H

Haiku

Short, concise poem

  • Fastest response time
  • Lowest cost
  • Surprisingly capable for size
  • Ideal for simple tasks & interfaces

Evolving Capabilities

Programming Improvement (SBench) 3% → 50% in 10 months

Claude 3.5 Sonnet reached 50% on SBench, a benchmark of real-world software engineering tasks

Claude 3.5 vs Claude 3

"Sonnet 3.5 increased its intelligence to the point where it was smarter than the original Opus 3 Model, especially for code but also just in general." - Dario Amodei

Versioning Challenges

"Naming is actually an interesting challenge... It's not like software where you can say 'oh this is 3.7, this is 3.8.' You have models with different tradeoffs."

3/10