Claude Glossary
Claude is the family name for the large language models produced by Anthropic. The name is a deliberate reference to Claude Shannon, the American mathematician and electrical engineer (1916–2001) whose 1948 paper A Mathematical Theory of Communication founded information theory. Anthropic’s choice of the name signals an intellectual lineage running through the technical-research tradition of communication and signal processing, in contrast to the corporate or mythological naming choices of peer models.
The Claude family ships in three tiers, each optimised for a different point on the capability/cost/latency curve:
- Claude Opus — the most-capable tier; best for hard reasoning, long-form careful writing, and complex agentic tasks.
- Claude Sonnet — the practical workhorse; balanced capability and cost; the default for most production agentic workloads.
- Claude Haiku — the fast and cheap tier; high-volume routine work, interactive latency-sensitive tasks.
Major versions to date: Claude 1 (March 2023), Claude 2 (July 2023), Claude 3 (March 2024), Claude 3.5 (June 2024), Claude 4 (October 2025), Claude 4.5 (December 2025), Claude 4.6 (February 2026), Claude 4.7 (April 2026).
For this Dictionary, Claude is the daily working model family. Most entries in this corpus have been drafted with Claude as the cooperative-writing collaborator, with the operator’s voice and editorial judgement as the final layer. The dependency is structural; the trade-offs are named in Opus Addict.
See also
- Anthropic — the producer
- Claude Opus, Claude Sonnet, Claude Haiku — the three tiers
- Opus Addict
- Constitutional AI — Anthropic’s training approach