Every AI term, model, and tool — explained clearly.
The definitive reference for artificial intelligence. Real examples, cited sources, weekly updates, and cross-references that show you how everything connects.
Explore AI, organized the way it actually works
Every term categorized so you can find neighbors, prerequisites, and next steps — not just a definition.
Foundational ML
Neural networks, gradient descent, backprop, loss functions, activation functions.
80 termsLLMs & Language Models
Context windows, tokens, prompting, temperature, system messages, streaming.
100 termsDeep Learning Architectures
Transformer, CNN, RNN, Mamba — the building blocks behind every model.
60 termsVector Databases & RAG
Embeddings, chunking, semantic search, rerankers, hybrid retrieval.
70 termsFine-tuning Methods
LoRA, QLoRA, DPO, GRPO, PPO, RLHF — every way to adapt a model.
50 termsAI Agents
MCP, tool use, ReAct, agentic loops, multi-agent orchestration.
50 termsMultimodal & Vision
Diffusion models, CLIP, VLMs, image and video generation architectures.
50 termsAI Safety & Alignment
Constitutional AI, red teaming, interpretability, jailbreaks, superalignment.
40 termsPrompt Engineering
Chain-of-thought, few-shot, ReAct, DSPy, prompt injection defenses.
40 termsEvaluation & Benchmarks
MMLU, HumanEval, GSM8K, MT-Bench — what benchmarks actually measure.
40 termsTraining & Optimization
ZeRO, FSDP, gradient checkpointing, mixed precision, Adam variants.
70 termsInference & Serving
vLLM, PagedAttention, KV cache, speculative decoding, prompt caching.
70 termsModern Architectures
Mamba, State Space Models, RWKV, Retentive Networks — beyond Transformers.
40 termsMixture of Experts
Sparse MoE, Switch Transformer, expert routing, DeepSeekMoE.
20 termsFoundation Model Families
GPT, Claude, Gemini, Llama, Mistral — the families reshaping AI.
60 termsReinforcement Learning
PPO, DQN, MCTS, policy gradients, AlphaZero-style training.
30 termsSpeech & Audio
Whisper, TTS, voice cloning, Suno, Udio — audio generation and understanding.
25 termsDatasets, Infra & Companies
The Pile, FineWeb, H100, TPU, HBM — what powers modern AI.
110 termsFlagship terms every AI practitioner should know
Deep explanations with real examples, historical context, and cross-references. Click copy on any definition to grab it for your notes.
Retrieval-Augmented Generation (RAG)
A technique that gives a language model access to an external knowledge source at inference time by retrieving relevant documents and injecting them into the prompt — reducing hallucinations and letting models answer from up-to-date data.
Read the full entry →Transformer
The deep learning architecture built around attention that processes sequences in parallel — foundation of GPT, Claude, Gemini, and nearly every modern large language model.
Read the full entry →Model Context Protocol (MCP)
An open standard introduced by Anthropic that lets AI applications connect to external data sources and tools through a uniform interface — the "USB-C for AI," now supported across major providers.
Read the full entry →Extended Thinking
A capability where a language model produces internal reasoning tokens before its final answer, working through multi-step problems more accurately — used by Claude, OpenAI o-series, DeepSeek R1, and others.
Read the full entry →LoRA (Low-Rank Adaptation)
A parameter-efficient fine-tuning method that freezes base weights and inserts small trainable low-rank matrices — cutting compute and memory cost by orders of magnitude.
Read the full entry →Mixture of Experts (MoE)
A neural network architecture where a router activates only a subset of specialized "expert" sub-networks per input — letting models scale to trillions of parameters while keeping inference cost tied to the active experts.
Read the full entry →Specs, benchmarks, pricing — updated weekly
Every major model with real specs, cited benchmarks, current pricing, and side-by-side comparisons. Because AI moves fast, we refresh these pages every week.
Claude Opus 4.8
Anthropic's flagship frontier model with extended thinking, industry-leading coding, and strong agentic tool use.
GPT-5
OpenAI's flagship model combining reasoning, multimodal understanding, and long-context tool use in a unified interface.
Gemini 3
Google's latest frontier model with strong long-context, native multimodal input, and deep Google-ecosystem integration.
Claude Sonnet 4.6
The workhorse balance-of-capability-and-cost model. Strong on coding, reasoning, and reliable tool use for production apps.
Llama 4
Meta's open-weight family with multiple sizes, competitive benchmarks, and permissive licensing for self-hosting.
Claude Haiku 4.5
Fast, low-cost Claude tier for high-volume applications, chat interfaces, and latency-sensitive agent loops.
Every AI tool, reviewed and compared
Honest reviews with pros, cons, alternatives, and what each tool is actually good for. No affiliate hype.
ChatGPT
OpenAI's flagship consumer chat interface with voice, vision, and custom GPTs.
Claude.ai
Anthropic's chat interface with Projects, Artifacts, Skills, and computer use.
Cursor
AI-native code editor built on VS Code with deep model-in-the-loop editing.
Perplexity
AI answer engine that cites sources — a search + LLM hybrid for research.
Midjourney
Image generation model known for strong aesthetics and stylistic consistency.
GitHub Copilot
The original AI code assistant — deep GitHub integration and IDE plug-ins.
Claude Code
Anthropic's agentic coding tool that runs in your terminal, VS Code, or JetBrains.
NotebookLM
Google's research notebook — turn your sources into summaries, podcasts, and Q&A.
Comparisons that answer "which one?"
Real benchmarks, real trade-offs, real decisions. Every comparison ends with a clear "which to choose when."
Claude Opus 4.8 vs GPT-5
Frontier flagship models compared on coding, reasoning, and cost.
RAG vs Fine-tuning
When to retrieve vs. when to adapt weights — with hybrid pattern.
Gemini 3 vs Claude Sonnet 4.6
Long-context multimodal vs. balanced workhorse for production apps.
LoRA vs QLoRA
Two parameter-efficient fine-tuning methods, memory and quality trade-offs.
Anthropic vs OpenAI
Two frontier labs compared on models, pricing, ecosystem, and philosophy.
MCP vs Function Calling
Two ways to give models tool access — when each is the right choice.
Prompt Engineering vs Fine-tuning
Reach for prompting first — but here's when fine-tuning genuinely wins.
Cursor vs GitHub Copilot
Two AI coding tools compared on editing model, cost, and workflow fit.
How things actually work — end to end
Long-form tutorials for concepts that need more than a definition. Diagrams, worked examples, and links to primary sources throughout.
How Transformers Work
Attention, embeddings, positional encoding — walked through step by step.
How RAG Works (End-to-End)
Chunking, embeddings, retrieval, reranking, generation — with code.
How Agentic AI Works
The agentic loop, tool use, planning, and orchestration patterns.
How Fine-Tuning Works
From SFT through RLHF, DPO, and GRPO — every method explained.
Understanding Context Windows
Why longer isn't always better and what "lost in the middle" really means.
How MCP Works
Servers, clients, transports — the protocol behind the AI plumbing standard.
Learning paths, from zero to expert
Ordered lessons that build on each other. Perfect when you don't know what you don't know yet.
AI from Zero
20 lessons taking you from "what is a neural network" to "I can read an AI paper."
IntermediateAI for Developers
30 lessons on integrating LLMs into real applications — APIs, RAG, agents, evals.
BusinessAI for Executives
25 lessons on AI strategy without the math — vendors, budgets, and org design.
SpecialistMaster RAG in 10 Lessons
Deep-dive into every RAG component — from chunking strategies to production evals.
AI, organized around what you actually do
Best tools, common workflows, prompt templates, and real case studies for your role.
AI for Software Engineers
Best coding assistants, agents, and testing workflows.
AI for Marketers
Content, research, and creative tools that actually save time.
AI for Lawyers
Contract review, research, and drafting — with the ethical guardrails.
AI for Product Managers
Research synthesis, spec drafting, and evaluating AI vendors.
AI for Teachers
Lesson plans, assessments, and student-facing tools done right.
AI for Researchers
Literature review, coding assistants, and writing tools for academic work.
Questions people ask about AI Terms Guide
AI Terms Guide is a comprehensive reference encyclopedia covering every artificial intelligence term, model, tool, provider, and concept. Each entry is written with real examples, cited sources, and deep cross-references. New terms are added within 48 hours of major releases, and pricing plus benchmarks are refreshed weekly.
Wikipedia is comprehensive but slow to update and shallow on AI-specific detail. AI Terms Guide focuses exclusively on AI, updates weekly, includes real code examples, cross-references models to concepts, offers side-by-side comparisons, and provides structured learning paths — none of which Wikipedia does for AI in a unified way.
Pricing and new model releases are updated weekly. Benchmarks are refreshed monthly. Every term page has a visible last-updated date. Major AI concepts get new pages within 48 hours of a paper or product announcement — see our update policy for details.
Content is written by our editorial team — Ahmed R. and Sana K. lead writing, and each technical page is reviewed by an AI engineer before publication. Every claim links to a primary source: a research paper, official documentation, or a trusted benchmark. See our editorial policy.
Yes — every term, model, tool, provider, and comparison page is free to read. Premium downloadable cheatsheets, comprehensive PDF glossaries, and enterprise embed licenses are available separately for those who need them.