AI Common Generative AI Terms

Table of Contents +

In the last lesson, you learned AI Cloud vs Local Models. Now let us make a quick glossary. A glossary is just a list of important words with simple meanings.

Here is how the common terms connect:

Prompt

Tokens

Context window

LLM

RAG

Tools and MCP

🤔 Why Keep a Glossary?

AI terms repeat everywhere.

  • Model docs use them: Documentation may say token, context, embedding, or inference without explaining every time.
  • API docs use them: Provider APIs use these words in request fields, pricing pages, and error messages.
  • Product teams use them: People building AI features discuss prompts, tools, memory, and evaluation.
  • Interviewers use them: AI and system design interviews may ask what RAG, tokens, or hallucinations mean.
  • Error messages use them: When a request fails, the error may mention context limit, max tokens, or invalid tool call.

So this page is your quick reference. Come back to it whenever a word feels unclear.

Do not try to memorize everything in one reading.

Prompt

The text you send to the model, such as a question, task, pasted document, or instruction.

Token

A small text piece the model reads or writes.

Context

The information available inside one model request.

  • Read the term: First see the word without fear.
  • Read the simple meaning: Connect it to one plain sentence.
  • Look at where it appears: Notice which lesson uses that term in a real flow.
  • Then continue: Repetition across lessons will make the word natural.

📖 How to Read AI Documentation

When you read model or API docs, slow down around these terms.

  • If the docs say tokens, think cost, limits, and length.
  • If the docs say context, think what information the model can see.
  • If the docs say tool, think external function or API call.
  • If the docs say embedding, think search and similarity.
  • If the docs say evaluation, think testing model behavior.

This habit makes documentation much easier to understand.

🌍 Terms in Real Sentences

Prompt

"Summarize this email" typed into ChatGPT, Claude, or Gemini is a prompt.

Model

A chat app can be built on a model family like GPT, Claude, Gemini, Llama, or Gemma.

Context

If the app sends a pasted document, old chat messages, or a support article with your question, that extra text becomes context.

📘 Common Generative AI Terms

Term Simple meaning
LLM Large Language Model; a model that reads and generates text
Token A piece of text processed by the model
Prompt The input instructions and context sent to the model
Context window The token-limited space available in one request
Embedding A numeric representation of text or other data
RAG Retrieval-Augmented Generation; retrieve data, then generate from it
Agent An AI workflow that can plan steps and use tools
Tool An external capability the AI app can call, like search or a database lookup
MCP Model Context Protocol; a standard way to connect AI apps with tools, resources, and prompts
Fine-tuning Additional training to specialize model behavior
Inference Using a trained model for a request
Hallucination Incorrect generated information presented confidently

⚖️ Terms That Often Get Confused

  • Prompt vs context: The prompt is the full input you send. Context is the information available inside that input.
  • RAG vs fine-tuning: RAG adds information at request time. Fine-tuning changes model behavior through extra training.
  • Tool vs model: The model generates or reasons. A tool fetches data or performs an action.
  • Agent vs chatbot: A chatbot replies. An agent may plan, call tools, and continue across steps.

🧪 Quick Developer Examples

Here is how these terms appear in code and architecture discussions.

  • “We need fewer tokens” means the request is too large or expensive.
  • “Add this to context” means include it in the model request.
  • “Use RAG” means retrieve relevant data before asking the model.
  • “Add a tool” means let the AI workflow call a real function.
  • “Run evaluation” means test outputs against expected behavior.

🧩 What You’ve Learned

  • ✅ Terms like LLM, token, prompt, context window, and RAG appear everywhere in AI apps.
  • ✅ Clear vocabulary helps you understand docs, APIs, and product discussions.

Check Your Knowledge

 4 questions Show quiz Hide quiz

Test what you learned. Pick an answer for each question, then click Check.

  1. 1

    What does RAG stand for?

    Why: RAG means retrieving relevant information and using it during generation.

  2. 2

    What is an embedding?

    Why: Embeddings represent text or other data as numbers for search and comparison.

  3. 3

    What is the difference between a tool and a model?

    Why: Tools extend the AI app with external capabilities.

  4. 4

    What does hallucination mean?

    Why: Hallucination means confident output that is not actually correct.

Score: 0 / 4

🚀 What’s Next?

When we build even a small AI app, these terms start connecting: prompt, backend, model, RAG, agent, MCP, and response. Next, let us build that small LLM app flow.

Previous AI Cloud vs Local Models Next AI Build Your First LLM Application