AI Glossary

A

• Models & technology AI Agent An AI agent plans and performs multiple steps toward a goal. Unlike a chatbot that only answers, an agent can use tools and take actions. Read the explanation →
• Models & technology App Intents App Intents is Apple's framework for exposing an app's actions and data to Siri, Shortcuts, Spotlight, and Apple Intelligence. Read the explanation →
• Models & technology Apple Foundation Models Apple Foundation Models are the family of models behind Apple Intelligence, split between models that run on the device and models that run in Private Cloud Compute. Read the explanation →

C

• Models & technology Claude Fable 5 Claude Fable 5 is Anthropic's top general-availability model, a Mythos-class system aimed at long, complex work and at turning vague requests into a design. Read the explanation →
• Models & technology Claude Opus 4.8 Claude Opus 4.8 is Anthropic's high-capability model for coding, agents, and professional knowledge work — the practical choice for demanding everyday tasks. Read the explanation →
• Models & technology Computer Use Computer use lets an AI look at a screen and click, type, and scroll — operating browsers and applications the way a person would. Read the explanation →
• Models & technology Constitutional AI Constitutional AI gives a model an explicit set of principles and has it critique and revise its own answers against them. Anthropic developed the approach. Read the explanation →

D

• Models & technology Deep Research Deep Research has an AI work through many sources autonomously and produce a cited research report. It compresses hours of background reading into minutes. Read the explanation →
• Models & technology Diffusion Model A diffusion model generates an image by starting from noise and removing it step by step. Most image generation AI works this way. Read the explanation →
• Models & technology Distillation Distillation transfers the capability of a large model into a smaller one, which is how capable AI ends up running on a phone. Read the explanation →

E

• Models & technology Embedding An embedding converts the meaning of words, text, or images into a sequence of numbers, so that closeness in meaning becomes something you can calculate. Read the explanation →

F

• Models & technology FLUX FLUX is a family of image generation and editing models from Black Forest Labs, usable through several services or run locally. Read the explanation →
• Models & technology Foundation Models framework The Foundation Models framework is Apple's developer API for using large language models from an app — summarization, generation, image understanding, and tool calling. Read the explanation →
• Models & technology Function Calling Function calling lets a model invoke external capabilities such as search or calculation when its own knowledge is not enough. It is what turns a chatbot into something that acts. Read the explanation →

G

• Models & technology GAN (Generative Adversarial Network) A GAN trains two networks against each other — one generating data, one judging it — until the generated output is indistinguishable from real data. Read the explanation →
• Models & technology GPT (Generative Pre-trained Transformer) GPT is OpenAI's series of large language models. It is the engine inside ChatGPT and the model family that set off the generative AI boom. Read the explanation →
• Models & technology GPT-5.6 GPT-5.6 is OpenAI's model family in three tiers — Sol at the top, Terra for everyday work, and Luna for fast, low-cost processing. Read the explanation →
• Models & technology GPTs GPTs let you build a customized ChatGPT for a specific purpose without code, and share it with others. Read the explanation →
• Models & technology Grounding Grounding ties an AI's answer to a verifiable source — search results, documents you supplied, or a database — in order to reduce hallucination. Read the explanation →

I

• Models & technology Image Generation AI Image generation AI produces illustrations and photographic images from a written description, without requiring design skills. Read the explanation →

K

• Models & technology Knowledge Graph A knowledge graph stores entities such as people, places, and products together with the relationships between them. It is used in AI search and in RAG. Read the explanation →

L

• Models & technology LoRA (Low-Rank Adaptation) LoRA fine-tunes a large model efficiently by training a small set of additional parameters instead of adjusting the whole model. Read the explanation →

M

• Models & technology MCP (Model Context Protocol) MCP is an open standard for connecting AI applications to tools and data. It gives different AI clients a common way to use compatible services. Read the explanation →
• Models & technology MLX MLX is Apple's open source framework for running machine learning efficiently on Apple silicon, widely used for local LLMs and agents on the Mac. Read the explanation →
• Models & technology Multi-Agent System A multi-agent system has several AI agents divide the roles and pass information and results between them to complete one goal. Read the explanation →

N

• Models & technology Nano Banana Nano Banana is Google DeepMind's family of Gemini image generation and editing models, ranging from a fast Lite tier to a quality-focused Pro tier. Read the explanation →
• Models & technology NLP (Natural Language Processing) NLP is the field of getting computers to understand, analyze, and generate human language. It underlies translation, search, and chat assistants. Read the explanation →
• Models & technology NPU (Neural Processing Unit) An NPU is a processor dedicated to running AI workloads quickly at low power. It is what makes on-device AI practical on phones and laptops. Read the explanation →

O

• Models & technology OCR (Optical Character Recognition) OCR converts text inside an image or scanned document into editable, searchable text. AI made it accurate enough for handwriting and awkward layouts. Read the explanation →

P

• Models & technology Predictive AI Predictive AI learns patterns from past and present data to predict values, classifications, or likelihoods for new data. Read the explanation →
• Models & technology Private Cloud Compute Private Cloud Compute runs AI requests too heavy for the device on Apple silicon servers, designed so that personal data is not retained or accessible to Apple. Read the explanation →
• Models & technology Prompt Engineering Prompt engineering is the practice of designing and improving instructions so an AI produces a more useful, reliable, and correctly formatted result. Read the explanation →
• Models & technology Prompt Injection Prompt injection uses crafted input to override a model's standing instructions and make it behave in ways its developers did not intend. Read the explanation →

Q

• Models & technology Quantization Quantization compresses a model by deliberately reducing the numeric precision of its parameters. It is why local LLMs run on ordinary computers. Read the explanation →
• Models & technology Qwen Qwen is Alibaba's family of language and multimodal models, spanning text, images, audio, and code, with many released as open weights. Read the explanation →

R

• Models & technology RAG (Retrieval-Augmented Generation) RAG has a model retrieve relevant material before answering and write its response from that material. It is the standard way to give AI current, source-backed knowledge. Read the explanation →
• Models & technology Reasoning Model A reasoning model takes time to think in stages before answering, which raises accuracy on mathematics and other complex problems. Read the explanation →
• Models & technology RLHF (Reinforcement Learning from Human Feedback) RLHF trains a model against human preferences to improve how it responds. It is why assistants like ChatGPT feel helpful rather than merely knowledgeable. Read the explanation →

S

• Models & technology Semantic Search Semantic search finds information by closeness of meaning rather than matching keywords, so differently worded text with the same intent still surfaces. Read the explanation →
• Models & technology Sol, Terra, and Luna Sol, Terra, and Luna are the three capability tiers of GPT-5.6 — Sol for hard problems, Terra for everyday work, Luna for fast, high-volume processing. Read the explanation →
• Models & technology Speech Recognition Speech recognition converts spoken words into text. It underlies transcription, voice input, and the listening half of voice conversation with AI. Read the explanation →
• Models & technology System Prompt A system prompt is the standing instruction given to a model before a conversation begins. An assistant's persona and its rules come from it. Read the explanation →

T

• Models & technology Text-to-Speech (TTS) Text-to-speech generates human-sounding speech from written text. AI narration has reached the point of being hard to distinguish from a recording. Read the explanation →
• Models & technology Transformer The Transformer is the neural network architecture introduced in 2017. Its attention mechanism handles context well, and it underpins essentially every modern LLM. Read the explanation →

V

• Models & technology Vector Database A vector database stores large numbers of embeddings and finds the ones closest in meaning quickly. It is the retrieval layer underneath RAG. Read the explanation →
• Models & technology Vibe Coding Vibe coding means describing what you want in natural language and letting the AI write the code, judging by whether it works rather than reading every line. Read the explanation →
• Models & technology Video Generation AI Video generation AI produces moving footage from text or a single image. Output now approaches live-action realism and is changing how video gets made. Read the explanation →

W

• Models & technology Whisper Whisper is OpenAI's speech recognition model. Released openly and free to use, it became the default component inside transcription tools. Read the explanation →
• Models & technology World Model A world model is an AI's internal model of how an environment behaves, used to predict what happens next in response to time and action. Read the explanation →

Z

• Models & technology Z.ai and GLM Z.ai is a platform offering AI chat, models, and developer APIs. GLM is the family of language and multimodal models it runs on. Read the explanation →