Claude Glossary: Key Terms for Beginners
Plain-English definitions of the words you'll see everywhere in Claude — model, token, context window, prompt, project, artifact, and more.
New to Claude? Before anything else clicks, it helps to know the handful of words that show up all over the interface and every tutorial. Here's each one in plain English, with why it matters to you.
Core Terms
| Term | What it means | Why it matters |
|---|---|---|
| Model | The specific version of Claude answering you (e.g. Sonnet, Opus, Haiku). Each is tuned for a different balance of speed, cost, and depth. | Picking the right model is the difference between a fast cheap answer and a slow thorough one. |
| Token | The unit Claude reads and writes in — roughly ¾ of a word. "Hello there" is about 3 tokens. | Limits and API prices are measured in tokens, not words. |
| Context window | How much text Claude can "see" at once: your prompt + uploaded files + the conversation so far. Claude's window is 200,000 tokens (~500 pages). | Everything you want Claude to reason over must fit inside it. |
| Prompt | The message or instruction you send Claude. | A clear prompt is the single biggest lever on answer quality. |
| Conversation / Chat | One continuous thread. Claude remembers everything within a chat, but starts fresh in a new one. | Keep related work in one chat so Claude keeps the context. |
Features You'll See in the App
| Term | What it means | Why it matters |
|---|---|---|
| Custom Instructions | Standing rules you set once (a role, tone, or format) that Claude follows in every chat of a Project. | Stops you from re-explaining how you want answers each time. |
| Project | A workspace that bundles a persistent knowledge base + custom instructions, so Claude already knows your context. | Best for anything ongoing — a codebase, a client, a book. |
| Knowledge base | The files you upload into a Project for Claude to reference. | Grounds answers in your material instead of guesswork. |
| Artifact | Generated content — code, a document, a diagram — shown in a live side panel you can edit and iterate on. | Turns Claude from a chat into a workspace you build in. |
| Vision | Claude's ability to read images: screenshots, charts, diagrams, handwritten notes. | You can paste a picture, not just type. |
Good to Know
| Term | What it means | Why it matters |
|---|---|---|
| Hallucination | When an AI states something false with confidence. Claude is designed to do this less and to admit uncertainty. | Always sanity-check facts that matter, on any AI. |
| Usage limit | A cap on how much you can send in a given period — Claude has shorter session limits and a weekly cap. Higher plans raise them. | If you hit a limit, it resets — or you upgrade for more headroom. |
| System prompt | Behind-the-scenes instructions that shape how a model behaves. In the app, your Custom Instructions play this role. | It's why the same model can act like an editor in one Project and a coder in another. |
That's the vocabulary you'll need. Next, see how the models differ and when to use each — or jump straight to the full Claude overview.