ChatGPT Glossary: Key Terms for Beginners
Plain-English definitions of the words you'll see across ChatGPT — model, token, custom instructions, GPTs, memory, and more.
ChatGPT throws a lot of terms at newcomers. Here are the ones that actually matter, in plain English, with why each is useful to you.
Core Terms
| Term | What it means | Why it matters |
|---|---|---|
| Model | The version of ChatGPT answering you. As of 2026 it runs on the GPT-5 family, with a lighter free model and more capable paid ones. | More capable models give better answers but may be slower or paid-only. |
| Token | The unit the model reads and writes in — roughly ¾ of a word. | Limits and API prices are counted in tokens, not words. |
| Context window | How much text the model can consider at once: your prompt plus the conversation so far. | Very long chats can push older messages out of view. |
| Prompt | The message you send ChatGPT. | A clear prompt is the biggest lever on answer quality. |
Features You'll See
| Term | What it means | Why it matters |
|---|---|---|
| Custom Instructions | Standing preferences (about you and how you want replies) applied to every chat. | Set your tone and context once instead of repeating it. |
| Custom GPTs (GPTs) | Mini versions of ChatGPT you configure with a persona, instructions, and knowledge, then reuse. | Build a specialized assistant once and use it repeatedly. |
| Memory | ChatGPT remembering useful facts about you across separate chats. | It can carry context between conversations, not just within one. |
| Advanced Voice | A natural, spoken back-and-forth conversation mode. | Talk to ChatGPT instead of typing. |
| Image generation | Creating images from a text description (DALL·E). | Ask for a picture, not just words. |
Good to Know
| Term | What it means | Why it matters |
|---|---|---|
| Hallucination | When the AI states something false with confidence. | Always sanity-check facts that matter, on any AI. |
| System prompt | Hidden instructions that shape how the model behaves; your Custom Instructions feed into this. | It's why the same model can behave differently in different setups. |
That's the vocabulary you'll need. For what ChatGPT does, its plans, and how it compares, see the full ChatGPT overview.