Textbook of AI

Claude is the family of large language models developed by Anthropic since 2022 and released publicly from March 2023. The models are named for Claude Shannon, the founder of information theory, and trained using Anthropic's Constitutional AI alignment methodology that uses AI-generated feedback rather than purely human feedback for harmlessness training.

Major releases: Claude 1 (March 2023), Anthropic's initial public release; comparable to GPT-3.5. Claude 2 (July 2023), improved reasoning and longer context window (100K tokens). Claude 3 family (March 2024), Haiku, Sonnet and Opus tiers; Opus competitive with GPT-4 on most benchmarks. Claude 3.5 Sonnet (June 2024), substantial reasoning and coding improvements; fastest cheapest tier in the family. Claude 3.5 Haiku / Claude 3.7 Sonnet (late 2024 to early 2025). Claude 4 family (late 2025–2026), Sonnet 4, Sonnet 4.5, Opus 4 onwards, Haiku 4.5. Claude Opus 4.7 (1M context) (2026), million-token context; the model under which this encyclopedia is being written.

The Claude models have been adopted at scale by enterprises through the Anthropic API, Amazon Bedrock, Google Cloud Vertex AI and the Claude.ai consumer product. Claude Code, Anthropic's CLI for Claude, is the engineering interface most directly relevant to this textbook's authoring environment.

Claude has been positioned by Anthropic as the safety-conscious frontier model, with explicit refusal behaviours, transparent reasoning capabilities, and tooling support for production deployment that emphasises predictability and auditability.