Claude

Claude is a proprietary large language model (LLM) suite developed by Anthropic, grounded in a safety‑first architecture known as Constitutional AI, guiding responses with ethical principles rather than purely human feedback.

The Claude family includes models like Haiku (lightweight tasks), Sonnet (balanced performance), and Opus (most advanced reasoning and coding), capable of handling inputs with contexts up to 1 million tokens.

It solves business problems around long‑form content creation, code generation, document analysis, and AI agents (e.g., Claude Cowork accessing local files) with safety guardrails and interpretability built in.

• Constitutional AI framework: Embeds ethical alignment into responses, reducing harmful outputs while maintaining utility.
• High‑context window models (up to 1M tokens): Supports massive inputs for large document or codebase analysis, enabling deep workflows.
• Model tiering (Haiku, Sonnet, Opus): Users can choose models aligned to task complexity—optimizing cost and performance.
• AI agent capability (Claude Cowork / Code): Models act on local files or terminal environments to execute complex, multi‑step workflows with minimal user prompting.
• Enterprise connectors (e.g., Microsoft 365 integration): Uses Model Context Protocol to analyze emails, files, and Teams chats for insight extraction in secure enterprise contexts.

• Content creators & editors: Use Sonnet or Opus to draft, structure, and polish long‑form writing with consistent tone; voice cloning tools support personalized style workflows.
• Software developers & engineers: Opus and Claude Code enable reasoning over large codebases, multi‑step task execution, and pull‑request generation.
• Enterprise knowledge workers: Microsoft 365 integration via MCP allows Claude to analyze emails, documents, and team discussions, streamlining insight extraction.
• Security teams & researchers: Mythos‑class capabilities (e.g., in Fable 5) and Project Glasswing enable high‑power analysis under controlled access for cybersecurity use cases.

• Account and access: Users go to claude.ai or install desktop/mobile apps; Free, Pro, Max, Team, or Enterprise subscriptions determine model and feature availability.
• Prompting: Users submit natural language prompts conversationally; models reply with text, code, or analysis, adhering to Constitutional AI guidelines.
• Model selection: Subscription level dictates which model is available—Free uses Sonnet; higher tiers unlock Opus, long‑context, agents, and features like Claude Code/Cowork.
• Agent workflows: Claude Cowork can access folders or files; Claude Code integrates into terminal for code tasks; model executes multi‑step tasks autonomously.
• Enterprise integrations: With Model Context Protocol, Claude connects to SharePoint, Outlook, OneDrive, and Teams to ingest context and generate insights.
• Advanced access: For elite users, Fable 5 (Mythos‑derived) is available under high‑tier/subscription for powerful capabilities with safety layers.

Anthropic employs tiered subscription pricing. Plans include a Free‑forever tier, Pro, Max (5× and 20× usage), Team (standard and premium per‑seat), and custom Enterprise contracts.

Higher plans unlock more usage, advanced models (Opus), agent capabilities, and enterprise connectors. API use is billed separately on a per‑token basis across Sonnet, Haiku, Opus, with consistent output‑to‑input pricing ratios; volume‑based and caching discounts apply.