Claude 3.5 Opus

Ingredient list

1. #1Test-Time Compute Scaling

   Invented by Google DeepMind · 2024-08 · Velocity 1.6y

   “Claude 3.5 Opus uses longer reasoning chains for complex problems, allocating more inference compute.”

   reasoningmedium
2. #2Constitutional AI

   Invented by Anthropic · 2022-12 · Velocity 3y

   “Anthropic's core alignment approach is Constitutional AI, used across Claude models.”

   alignmenthigh
3. #3ReAct (Reason + Act)

   Invented by Princeton / Google · 2022-10 · Velocity 3y

   “Claude 3.5 Opus includes 'Artifacts' feature for interactive tool use and reasoning.”

   agentsmedium
4. #4Self-Consistency

   Invented by Google · 2022-03 · Velocity 4y

   “Anthropic mentions using 'majority voting' for reasoning tasks in Claude 3.5 Opus.”

   reasoningmedium
5. #5Chain-of-Thought Prompting

   Invented by Google · 2022-01 · Velocity 4y

   “Claude 3.5 Sonnet (same family) demonstrates chain-of-thought reasoning in technical report examples.”

   reasoninghigh
6. #6Instruction Tuning (FLAN)

   Invented by Google · 2021-09 · Velocity 5y

   “Claude models are instruction-tuned; Claude 3.5 Opus shows strong instruction following.”

   traininghigh
7. #7Transformer Self-Attention

   Invented by Google · 2017-06 · Velocity 9y

   “Claude models are transformer-based; Opus uses standard transformer architecture.”

   architecturehigh