Learn the language
AI glossary
Large language model (LLM)
A model trained on lots of text so it can predict and generate language.
Generative AI
AI that creates new text, images, audio, or video instead of only classifying existing data.
Prompt
The instructions or examples you give a model so it knows what to do.
Prompt engineering
The craft of writing instructions that steer a model toward a useful result.
System prompt
Hidden or setup instructions that steer how a model behaves across a conversation.
Hallucination
When a model states something false as if it were a fact.
Retrieval-augmented generation (RAG)
A pattern that lets a model look up your documents before it answers.
Embedding
A numeric representation of text or other media used for search and similarity.
Vector database
A store built to search embeddings by similarity instead of exact keywords.
Token
A small chunk of text (or other data) that a model reads and writes.
Context window
How much material a model can consider at once in a single request.
Multimodal
A model or product that can work with more than one kind of input, such as text and images.
AI agent
Software that can take a goal and use tools or steps, not only reply with text.
Copilot
An assistant that sits inside an app you already use and helps with the task at hand.
Tool use (function calling)
When a model can call an external function, API, or app instead of only writing text.
Fine-tuning
Extra training that adapts a base model to a narrower style or task.
LoRA
A lighter way to fine-tune a model by training a small set of extra weights.
Diffusion model
A common architecture for generating images (and some video) by iteratively denoising.
Text-to-image
Generating a still image from a written prompt.
Text-to-video
Generating moving images from a prompt or a still.
Text-to-speech (TTS)
Turning written text into spoken audio.
Speech-to-text (transcription)
Turning spoken audio into a written transcript.
Temperature
A setting that controls how deterministic or varied a model’s next tokens are.
Few-shot prompting
Including a few examples in the prompt so the model can copy the pattern.
Inference
The step where a trained model produces an output for a new input.
API
A programmed interface so your software can call a model or a tool.
Open-source (or open-weight) model
A model whose weights, and sometimes code, are released for others to run.
OCR (optical character recognition)
Software that reads text out of an image or a scanned page.
AI watermarking
A signal added to generated media so it can later be identified as synthetic.
Custom GPTs
Task-specific ChatGPT setups with extra instructions and tools.
MCP (Model Context Protocol)
An open protocol for connecting assistants to external tools and data.
Alignment
The work of making a model behave in line with human goals and policies.
Top-p (nucleus sampling)
Another generation setting that limits the model to the most likely next tokens.
Computer use
A model that can operate a computer interface — clicking, typing, browsing — on your behalf.
Freemium
A pricing model with a useful free tier and paid upgrades.
Directory last reviewed . Confirm current prices on official sites.