Claude 1.0
Anthropic · March 2023
Why It Matters
First commercial model built with Constitutional AI, demonstrating that safety-focused training could produce a competitive assistant. Established Anthropic as a credible alternative to OpenAI.
Description
Anthropic's first publicly available AI assistant. Trained using Constitutional AI — a technique where the model learns safety guidelines from a set of written principles rather than relying solely on human feedback. Launched with a 9K token context window and a lighter 'Instant' variant.
Key Innovations
Family Tree
Successors (1)
Related Research (2)
Pioneered the RLHF paradigm — training a reward model from human preferences, then using it to fine-tune policies via reinforcement learning.
Introduced RL from AI Feedback using "constitutions" (rule sets) for self-supervision, reducing reliance on human labels for harmlessness training.