Learn AI
Short, plain-English explanations of the terms you meet when choosing AI tools and models - each linked to live examples from our database.
- AI agents
- An AI agent is software that uses an AI model to plan and carry out multi-step tasks by calling tools - running code, browsing websites or operating apps - instead of only replying with text.
- AI coding agents
- An AI coding agent reads and edits a codebase, runs commands and tests, and iterates on the result - going beyond code completion to carry out whole programming tasks.
- Computer-use and browser agents
- Computer-use and browser agents operate a real browser or desktop - clicking, typing and reading the screen - to complete tasks in apps that have no API.
- Large language models (LLMs)
- A large language model is a neural network trained on large amounts of text to predict the next token, which lets it write, summarise, translate, reason about and generate code.
- Reasoning models
- Reasoning models spend extra computation "thinking" through intermediate steps before answering, which improves results on maths, coding and multi-step problems at the cost of speed and tokens.
- Multimodal AI
- Multimodal AI models accept or produce more than one kind of data - for example text and images, or audio and video - rather than text alone.
- Open-weight models
- An open-weight model publishes its trained parameters, so anyone can download and run it on their own hardware - subject to its licence, which may restrict some uses.
- Context window
- The context window is the maximum amount of text - measured in tokens - that a model can take into account in one request, including the conversation, documents and its own reply.
- Tokens and AI API pricing
- AI APIs charge per token - small chunks of text - usually quoted in US dollars per million input tokens and per million output tokens, with output typically priced higher.
- AI benchmarks
- AI benchmarks are standard test sets used to measure model capabilities such as reasoning, maths or coding, so models can be compared on the same tasks.
- Epoch Capabilities Index (ECI)
- The Epoch Capabilities Index is Epoch AI's composite score of model capability, estimated from results across many benchmarks with a statistical model so that models can be placed on one scale.
- Model Context Protocol (MCP)
- The Model Context Protocol is an open standard for connecting AI applications to external tools and data sources through a common interface, so one integration works with many AI apps.
- Retrieval-augmented generation (RAG)
- Retrieval-augmented generation lets an AI model answer from your own documents: relevant passages are searched and retrieved first, then given to the model together with the question.
- AI image generators
- AI image generators create pictures from a text description (text-to-image), and many can also edit existing images - extending them, removing objects or changing styles.